The Cuban missile crisis and intelligence performance

This article was downloaded by: [Harvard College] 

On: 18 September 2012, At: 08:23 

Publisher: Routledge 

Informa Ltd Registered in England and Wales Registered Number: 

1072954 Registered office: Mortimer House, 37-41 Mortimer Street, 

London W1T 3JH, UK 

Intelligence and National 

Security 

Publication details, including instructions for 

authors and subscription information: 

http://www.tandfonline.com/loi/fint20 

The Cuban missile 

crisis and intelligence 

performance 

James G. Blight a & David A. Welch b 

a Professor of International Relations 

(Research) at the Thomas J. Watson Jr 

Institute for International Studies, Brown 

University, USA 

b Associate Professor of Political Science, 

University of Toronto 

Version of record first published: 02 Jan 2008. 

To cite this article: James G. Blight & David A. Welch (1998): The Cuban 

missile crisis and intelligence performance, Intelligence and National Security, 

13:3, 173-217 

To link to this article: http://dx.doi.org/10.1080/02684529808432498 

PLEASE SCROLL DOWN FOR ARTICLE 

Full terms and conditions of use: http://www.tandfonline.com/page/ 

terms-and-conditions 

This article may be used for research, teaching, and private study 

purposes. Any substantial or systematic reproduction, redistribution,

Downloaded by [Harvard College] at 08:23 18 September 2012 

reselling, loan, sub-licensing, systematic supply, or distribution in any 

form to anyone is expressly forbidden. 

The publisher does not give any warranty express or implied or make 

any representation that the contents will be complete or accurate 

or up to date. The accuracy of any instructions, formulae, and drug 

doses should be independently verified with primary sources. The 

publisher shall not be liable for any loss, actions, claims, proceedings, 

demand, or costs or damages whatsoever or howsoever caused arising 

directly or indirectly in connection with or arising out of the use of this 

material.


The Cuban Missile Crisis and 

Intelligence Performance 

JAMES G. BLIGHT and DAVID A. WELCH 

The foregoing essays collectively raise important questions, not only about 

what happened in the Cuban missile crisis (and why), but also about 

intelligence assessment more broadly. 1 As we stated in our introduction, we 

see them as the beginning of a more extended conversation between 

scholars, intelligence professionals, and (we would hope) policy makers, 

rather than as the final word on any topic in particular. A proper synthesis 

is clearly premature. But given the Sisyphean nature of missile crisis 

scholarship, it may be doubted whether the time will ever be ripe for a true 

synthesis. Hence in this essay we make so bold as to offer an interpretation 

of the collective import of the preceding pages, knowing full well that some 

of it must later stand to be corrected. 

We would like to try to grapple with deep and abstract questions having 

to do with the evaluation of intelligence performance, and the theory and 

practice of intelligence assessment — though we shall evince some 

skepticism about the possibility of a 'theory' of intelligence assessment per 

se. While not everything we will have to say on this score will be entirely 

novel, we hope that the perspective in which we will attempt to put it will 

be unfamiliar and fresh. These are our chief aims. However, we believe it 

would be fruitful to ease our way into these topics by first having a closer 

look at some of the nuts and bolts of intelligence assessment in the missile 

crisis, both to provide some real-world purchase on more abstract issues, 

and to flag items that require further exploration. We begin with unresolved 

or puzzling empirical questions that arise from the preceding essays - 

questions first about what actually happened, and then about why certain 

things happened and others did not - and then proceed to examine and 

evaluate certain important perceptions, judgments, and inferences. This will 

help us frame and anchor the more abstract discussion to follow.


174 INTELLIGENCE AND THE CUBAN MISSILE CRISIS 

'WHAT' QUESTIONS 

Owing to the various dimensions of imbalance that we discussed in our 

introductory essay, rather more questions about what actually happened in 

intelligence assessment and intelligence-policy interactions in the Cuban 

missile crisis arise with respect to the Soviet Union and Cuba than with 

respect to the United States. This is not to say that the American intelligence 

story is an open book - far from it - merely that there are relatively few 

aspects of the American story that remain utterly mysterious to outsiders. 

The single most important lacuna, of course, is signals intelligence (Sigint), 

encompassing communications intelligence (Comint) and electronic 

intelligence (Elint). Whether the United States had broken Soviet and 

Cuban codes - and if so, what of importance, if anything, they learned from 

eavesdropping on Soviets and Cubans - are some of the important pieces of 

information not yet on the public record. Neither do we have details of 

Soviet or Cuban Sigint, of course - but it is at least ironic that in the Soviet 

and Cuban cases, we have clear testimony to the effect that their capabilities 

were limited and not very useful. 2 We do not even have a general 

characterization of this sort with respect to American Sigint. As Raymond 

Garthoff notes, one direct reference to signals intelligence in a nowdeclassified 

document, coupled with judicious inferences about security 

deletions from other declassified documents, permits the surmise that Sigint 

was on balance more useful to the United States than was human 

intelligence (Humint), but less useful than photointelligence (Photint). 3 This 

would confirm the sense students of the missile crisis without security 

clearances have had (ourselves included) that the United States probably did 

benefit from signals intelligence, but that our understanding of the event 

would not change radically if the Sigint story could be told in full. But this, 

of course, remains to be seen. 

Although Aleksandr Fursenko and Timothy Naftali have done a 

remarkable job of painting what was until very recently an almost entirely 

blank canvas, the Soviet intelligence picture is less complete. Three 

question in particular arise as a result of apparent inconsistencies between 

the accounts of Fursenko and Naftali on the one hand, and Domingo 

Amuchastegui on the other; a fourth question arises out of inconsistencies 

between the documentary and testimonial records. 

1. Did the KGB consider an American invasion of Cuba imminent between 

the Bay of Pigs and the Cuban missile crisis, and if so, did they report this 

to Moscow? Fursenko and Naftali maintain that the KGB provided the 

Kremlin with raw information that could be construed in two quite different 

ways: suggesting both that there was great danger of an American invasion


INTELLIGENCE PERFORMANCE 175 

of Cuba, and that such an invasion was unlikely. They suggest that the KGB 

hedged its reports, even at the price of blatant inconsistency, so as to be able 

to defend itself in the face of any contingency. 4 Nevertheless, the KGB was 

unable to reach a firm determination of American intentions toward Cuba, 

and never issued a clear warning that an invasion was imminent. The best 

and most insightful intelligence suggested that it was not. Aleksandr 

Feklisov, the KGB rezident in Washington, judged quite perceptively that 

Kennedy would have to be provoked into invading Cuba and would not 

willingly attack otherwise. 5 Fursenko and Naftali even imply that in judging 

an American attack unlikely, the KGB relied in part upon Cuban 

assessments. 6 Amuchastegui, however, reports that in 1961 Soviet 

intelligence began furnishing Cuba with 'constant, unambiguous, dire threat 

assessments ... for which there was no good-quality evidence'. 7 

As Amuchastegui readily acknowledges, the lack of Cuban documents 

represents a major obstacle for scholars who wish to get to the bottom of 

issues such as this. It is entirely possible, however, that Fursenko and 

Naftali on the one hand, and Amuchastegui on the other, are both correct. 

The KGB may have deliberately skewed its reports to Cuba for some 

ulterior purpose. Indeed, this is what Amuchastegui reports Cuban 

intelligence eventually decided. The conclusion Cuban intelligence 

retrospectively drew from this - that Khrushchev must have conceived the 

nuclear deployment in 1961, and that he made use of falsified KGB threat 

assessments to prime Castro to accept it - is logically consistent with, 

though not logically entailed by, such dissimulation. Khrushchev may have 

had other goals in mind in pursuit of which it would have been useful to 

persuade Castro that the Cuban Revolution was in imminent danger from 

the United States. He might simply have sought to draw Cuba closer to the 

Soviet Union, for instance. 8 

2. What did the Soviet intelligence community know about the plan to 

deploy missiles, and when did it know it? Fursenko and Naftali claim that 

Aleksandr Alekseev was the only KGB field officer who was briefed on the 

Cuban initiative. 9 Alekseev was summoned to Moscow in May, appointed 

ambassador, informed of the plan, and dispatched back to Cuba with a 

delegation led by alternate Presidium member Sharaf Rashidov and Deputy 

Minister of Defense (and Commander-in-Chief of the Strategic Rocket 

Forces) Marshal S. S. Biryuzov, whose primary task was to sell the idea to 

Castro. 10 By all accounts, however, neither Georgi Bolshakov - the GRU's 

man in Washington — nor Aleksandr Feklisov knew of the deployment 

before President Kennedy announced it to the world on 22 October. 

Semichastny claims he found out about the deployment only when the KGB 

took custody of nuclear warheads aboard Soviet merchant ships in late



summer." This implies that certain KGB units charged with transport and 

custody of nuclear munitions must also have been aware of the decision at 

some point. 12 But without knowing precisely who in the Soviet intelligence 

community knew of the deployment, when they knew it, and why they were 

(or were not) informed, it is difficult to come to a refined assessment of the 

relevance and performance of Soviet intelligence in the episode. 

3. Did the KGB orchestrate a campaign to mask the nuclear deployment by 

funneling accurate information about it to the CIA through Cuban sources? 

If the KGB was unaware of the missile deployment, then it could not have 

attempted to mask it by leaking accurate information about it. Yet Domingo 

Amuchastegui claims that this is precisely what the KGB did: 'This 

campaign presupposed - quite correctly - that the CIA would discount this 

information, because they would not consider the individuals and groups 

peddling it to be credible." 3 One can only admire the daring and the genius 

of such a plan. But if Fursenko and Naftali are correct that Alekseev was the 

only KGB field officer briefed on the deployment, then unless Alekseev 

personally directed it - and he has never claimed that he did - 

Amuchastegui must be mistaken. 

Here, again, we must await further documentation for a definitive 

resolution. But it is entirely plausible to imagine that the KGB and Cuban 

intelligence did cooperate on such a campaign. 

One possibility is that the KGB in Cuba did, in fact, know about the 

nuclear deployment. Indeed, it is difficult to imagine that they did not. 

Thousands of Soviet troops in Cuba surely were aware of it; why would the 

local KGB be unaware? If they had not been informed in advance, certainly 

they would have discovered it in any case (Cuban intelligence did, after all, 

despite not being informed). But it is difficult to imagine why the local KGB 

would not have been informed. They would have been the most useful cadre 

imaginable for protecting the secrecy of the deployment. It is easy to 

imagine that Fursenko and Naftali err to imply that only Alekseev knew. 

It is equally easy to imagine, however, that Amuchastegui errs in 

recalling - or perhaps in assuming - that the KGB could only have 

conceived and carried out such a plan if they knew that the deployment 

actually included strategic nuclear missiles. The KGB officers involved in 

designing the campaign could have believed that the Soviet Union was only 

deploying conventional military forces to the island, and yet still would 

have appreciated the value of tales about nuclear missiles. Their task was to 

capitalize on the CIA's low estimate of Cuban credibility. Why would they 

not make use of stories that they, in their ignorance, believed to be false, and 

as such all the more incredible? 14



4. What exactly happened between Scali and Feklisov? This particular 

question is shaping up to become one of the enduring mysteries of the 

Cuban missile crisis. Fursenko and Naftali do an excellent job of 

summarizing what we know, and sketch out plausible scenarios. 15 We fully 

concur with their judgments, and have nothing to add to their discussion. 

We would like, however, to highlight their conclusion, which underscores 

the truly important point: 'While it is only speculation that the contradiction 

between the accounts suggests that Scali and Feklisov misunderstood each 

other on 26 October, it is certain that the United States and the USSR 

misunderstood what went on between the two men." 6 Any positive effect 

this intelligence side-show may have had was entirely fortuitous. 

Two additional 'what' questions arise from Domingo Amuchastegui's 

essay, one of which has important implications for the accuracy of the 

historical record, and one of which does not. The minor issue is whether 

Cuba accepted Soviet offers of intelligence assistance, as Fursenko and 

Naftali state, or whether Cuba requested it, as Amuchastegui implies. 17 

Nothing of real importance would seem to turn on the answer, though the 

divergent representations may be evidence of the turbulent undercurrent of 

national pride that undoubtedly complicated the Soviet-Cuban intelligence 

relationship, just as it profoundly complicated the Soviet-Cuban political 

relationship. 18 

The important question is whether Che Guevara traveled to Moscow in 

April 1962. Amuchastegui provides a detailed account of the mission, and 

characterizes it as a crucial turning point in Soviet-Cuban relations. 

According to Amuchastegui, the purpose of the trip was not only to request 

increased levels of military aid, but also military aid of a 'qualitatively 

different' character. Amuchastegui claims that Che 'left it up to IChrushchev 

himself to make a concrete proposal', and that '[a]t no point did Che suggest 

a deployment of Soviet nuclear missiles, nor did Khrushchev offer them at 

that time'. 19 Nevertheless, if Khrushchev understood Che to be probing for 

an offer of nuclear assistance, this could certainly help explain why the idea 

took root in Khrushchev's own mind almost immediately thereafter. The 

timing would be perfect. 

The difficulty is that we have been unable to confirm Amuchastegui's 

claim that any such trip took place. We cannot locate any reference to such 

a mission in any of the published literature, and the knowledgeable 

Cubanologists whom we have consulted on the matter have no recollection 

of it. Neither Alekseev nor Oleg Daroussenkov - Che's special assistant in 

1961 and 1962, a former staff member and interpreter in the Soviet embassy 

in Havana, and thereafter the leading Cuba specialist at the Central 

Committee - can remember such a visit. Nor is there any evidence of it in



the material Aleksandr Fursenko viewed in the Russian Presidential and 

Foreign Intelligence Service (SVR) archives during the course of his 

collaboration with Timothy Naftali. 20 

It is not out of the question that Che undertook such a mission, but it is 

highly curious that this particular trip would be shrouded in such secrecy 

when other trips to the Soviet Union by high-ranking Cubans dealing with 

equally sensitive topics (arguably, more sensitive topics) have been matters 

of public record for some time. This would suggest that if Che did travel to 

Moscow in April, something far more sensitive than Amuchastegui 

suggestions must have transpired - something of the order of an explicit 

request for nuclear assistance. By all accounts, however, Castro evinced 

surprise when Biryuzov suggested a nuclear deployment during his trip to 

Cuba at the end of May. 

Might Che have made such a request on his own initiative? Che and 

Castro certainly had their share of disagreements, but they were very close, 

and it is difficult to imagine that Che would make such a request without at 

least vetting it with Fidel. In any case, since the Soviets would not have 

known the purpose of Che's visit in advance, there would have been no 

reason why Alekseev or Daroussenkov would have been unaware at least of 

the fact that Che intended to go to Moscow. Nor would there have been any 

reason for the visit to pass entirely without mention in Soviet documents. If 

the trip indeed took place, then we must wonder very seriously about the 

recollections of former Soviet officials, the accuracy and completeness of 

the Russian archives that have been opened to scholars, or both. 

Is someone simply lying? We find it difficult to imagine what personal, 

professional, or national interest would be served by fabricating such a 

story. Nothing in it smacks of personal or professional accomplishment, let 

alone triumph. The story does not tend to cast Cuba, Cuban intelligence, 

Che Guevara, or Khrushchev in either a favorable or an unfavorable light. 

It provides no basis for explaining away an embarrassment, or escaping 

some blame. It is simply an interesting story with no particular valence. 

It is possible that Amuchastegui has some other trip in mind. There are 

two possibilities. Che did travel to Moscow in late 1960 for the purpose of 

requesting increasing levels of military aid, and there are at least some 

indications that, while in Moscow, he did float a trial balloon about a Soviet 

nuclear deployment in Cuba. 21 However, Amuchastegui distinctly connects 

Che's trip and the Escalante affair (March 1962), which means that if he 

were thinking of Che's 1960 trip, he would have to be confusing two 

significant events separated by almost a year and a half. Of course, 

Amuchastegui's claim that 'Khrushchev did not raise the Escalante affair or 

the expulsion of Kudriavtsev during the course of his meetings with Che' 22 

would certainly make sense if the trip had taken place in late 1960, rather



than in April 1962, since the Escalante affair had not yet taken place. 23 

A second possibility is that Amuchastegui is thinking instead of the trip 

to Moscow by Ramiro Valdes in March 1962. The timing is much better; but 

on first glance the context would appear to be wrong. Available records 

indicate that Valdes dealt primarily with KGB officials in Moscow, and that 

the chief purpose of his visit was to coordinate strategy for promoting 

revolution in Latin America (a task in which he failed utterly, much to his 

disgust). 24 However, Valdes was certainly a member of Castro's inner circle, 

and it is entirely possible that he could have been charged with the task of 

requesting increased military assistance. He could easily have approached 

Khrushchev secretly on this, and it would not be surprising that no mention 

of it appears in the available accounts. It may also be significant that when 

Rashidov and Biryuzov traveled to Cuba in late May to propose a nuclear 

deployment, Valdes was one of the five Cuban leaders (along with Fidel 

Castro, Raul Castro, Che Guevara, and Osvaldo Dorticos) who met to 

decide whether to accept the demarche. 25 This solution to the puzzle 

involves the fewest logical hurdles, the smallest number of facts which must 

be adjusted (only two: the date of the trip, and the lead member of the Cuban 

delegation), the fewest gaps in the documentary record that must be 

accounted for, and the smallest number of people whose recollections must 

be mistaken. Additionally, of the two facts that must be adjusted, one - the 

date of the trip - requires a trivially minor adjustment. 

While we must await further evidence from Cuban (or Russian) sources 

before we can solve this particular mystery, it is worth noting here what may 

hang on its outcome: namely, whether Cuba prompted, or Khrushchev 

proposed, a nuclear deployment. This is not a question that concerns 

intelligence specifically, but it is one that certainly bears on the genesis of 

the crisis. If Amuchastegui's story is correct, Castro may have deliberately 

sowed the seed that germinated in Khrushchev's mind. 

'WHY' QUESTIONS 

It is likely that as the documentary and testimonial record grows richer, 

many more questions about what actually happened will arise. For now, 

though, we wish to put them aside and ask why certain things happened, and 

why others did not. All five of the preceding essays do this throughout, of 

course, in somewhat different ways, and space prohibits a comprehensive 

review. Nevertheless, we would like to flag and explore some of the more 

interesting and more important puzzles before we move on to consider 

deeper issues, such as what constitutes a 'good answer' to questions of this 

kind. For the moment we are still concerned with fairly nitty-gritty matters 

of historical detail that emerge from the three empirical essays. Later in the



essay we will tackle the larger, more cosmic questions, such as why the 

three countries so badly misunderstood each other - questions which are the 

primary focus of the essays by Wirtz and Fischer. 

US Puzzles 

As we explain in our introductory essay, we believe the two most commonly 

asked 'why' questions about American intelligence in the Cuban missile 

crisis - (1) Why did DCI McCone, but not his analysts, anticipate a Soviet 

deployment?; and (2) Why did American intelligence discover the missiles 

only in mid-October? - are both relatively unimportant and fairly easy to 

answer when understood - as most historians of the crisis have understood 

them - as questions of historical detail. McCone got lucky (an insight 

devoid of practical implications for intelligence); the weather in Cuba was 

bad (ditto). We do not propose to consider them further here. They do, 

however, raise larger 'why' questions which we propose to examine in some 

detail later: namely, (1) Why did US intelligence, in attempting to estimate 

the likelihood of a Soviet nuclear deployment in Cuba, use as the basis of 

their judgment one set of assumptions about Soviet behavior rather than 

some other available set that might have led to a radically different 

estimate?; and (2) Why did American intelligence employ such an exacting 

standard of proof — direct photographic evidence of nuclear missiles — 

before being willing to conclude that a deployment of the kind President 

Kennedy had specifically warned the Soviet Union against, both publicly 

and privately, was, in fact, underway? 

To help answer these larger 'why' questions, it would be helpful to know 

the answers to two related nitty-gritty questions. First, why did US analysts 

ignore the estimate of the size of the Soviet force in Cuba provided by the 

operational branch of the CIA? Samuel Halpern has testified that Task Force 

W concluded in September 1962 that there were 45,000-50,000 Soviet 

military personnel in Cuba. The official American estimate in early October 

was 4,000-4,500 - an order of magnitude lower. 26 As we now know, the 

operations people had it just about right. Halpern is certain that CIA 

analysts received the Task Force W estimate in September, but does not 

know what happened to it. Why did the estimators dismiss it? 

The only defensible reason would be that they had some contrary 

evidence to suggest that the estimate must have been off by about an order 

of magnitude. If so, this evidence has never been made public. There are at 

least three other possible (non-mutually-exclusive) explanations - one 

bureaucratic, one cognitive, and one motivated: 

• The estimators may simply have been defending their turf. Estimates, 

after all, were their job.



• The CIA's presumption about the nature of the Soviet mission - to train 

Cuban troops - may have skewed their estimate. The CIA did not 

believe at that time that the Soviets intended to deploy to Cuba a 

substantial nuclear deterrent and a full battle-capable conventional force 

to defend the island from American attack. A training mission would 

have required no more than a few thousand Soviet troops, but many 

times more would be needed to mount a proper deterrent and defense. 27 

They may simply have seen what they expected to see. 

• CIA estimators may not have wanted to believe they were off by an 

order of magnitude, because of the painful personal or professsional 

repercussions of being wrong. 

We cannot say for certain which (if any) of these explains the error, singly 

or in combination, although our sense, from discussions with scholars and 

former intelligence professionals - as well as from a close reading between 

the lines of the available documents — is that the cognitive explanation is 

most plausible, the bureaucratic explanation somewhat less no, and the 

motivational explanation unlikely. 28 

The second question is related, and may be an important part of the 

answer to the first: Why did CIA estimators so heavily discount information 

from Cuban sources? Is Domingo Amuchastegui correct to infer that 'there 

was a pervasive and unprofessional distrust of Cuban sources"? 29 Garthoff 

writes: 

Most reports were of no value, or perhaps negative value; there were 

literally thousands of reports of missiles in Cuba in the period before 

any missiles were actually brought there. Only in late September and 

early October (after the 19 September SNIE) were a few reports from 

Opa-Locka received in Washington that, in retrospect, probably were 

valid sightings of the medium-range missiles — but that could not be 

determined at the time. The existence of hundreds of other reports that 

were found to be not valid, and many hundreds of others of 

undetermined validity, made it ever more difficult to credit the few 

that were later found to be true. 30 

It is one thing to question the accuracy and usefulness of a report because it 

is vague, confused, insufficiently detailed, self-contradictory, or 

inconsistent with other reliable information; it is entirely a different thing to 

question the accuracy and usefulness of a report because one suspects the 

motives, competence, or intelligence of the class of persons to which the 

reporter belongs. The poor quality or suspiciousness of one Cuban refugee 

report provides no valid logical ground for discounting another, high-



quality, Cuban refugee report; nor do 50 low-quality or suspicious reports 

provide twice the reason for discounting the 51st report as do 25 for 

discounting the 26th. Every single report should be assessed on its merits. 

We would certainly criticize a doctor who based a diagnosis on the 

frequency of diseases in the general population, rather than on the patient's 

symptom profile: the fact that a disease is extremely rare does not mean that 

any given patient must not have it. 

We do know that a few reports from Cuban sources were gems - very 

precise and very detailed descriptions of what, in retrospect, almost 

certainly were Soviet MRBMs. 31 We also know that some of these reports 

reached Washington well before the photographic discovery of missiles on 

15 October. We also know that, as a result of numerous false sightings and 

misinterpretations, 'CIA analysts had naturally come to view all such 

reports with a high degree of suspicion.' 32 It is easy to imagine that the CIA 

was prejudiced against Cuban sources; that this prejudice delayed the 

Kennedy administration's conclusion that a nuclear deployment was 

underway; and that it prevented the CIA from appreciating the true size of 

the Soviet expeditionary force. 

Soviet Puzzles 

Many questions arise with respect to Soviet intelligence, of which clearly 

the most important - stated most generally - surely must be, 'Why was 

intelligence so marginal to Soviet policy making?' 33 We can ask four more 

specific questions in an attempt to probe the possible answers. 

1. Why was the KGB not tasked to assess the likely US response to a nuclear 

missile deployment? Khrushchev had a deep and abiding fear of nuclear 

war, and apparently also genuinely worried that the United States might 

crush the delicate flower of socialism in Cuba. It is fair to assume that if 

Khrushchev truly believed that the United States, in response to a covert 

nuclear deployment, would take measures dramatically increasing the 

danger of both outcomes, he would not have undertaken it. He must 

therefore have believed that they would not. Khrushchev could not 

reasonably consider himself an expert on American politics, or on President 

Kennedy; so it would be natural - if he were reasonable - for him to ask 

those better qualified than himself to prepare a sober assessment. 

Among those better qualified was his personal aide on foreign policy 

matters, Oleg Troyanovsky, who, as the son of the first Soviet ambassador 

to the United States, had grown up in Washington, attended Swarthmore 

College, and knew the American political scene extremely well. 

Troyanovsky's testimony is worth quoting at some length:



Even though I was familiar with all the information that IChrushchev 

received on foreign policy, I did not immediately find out about his 

intention to deploy nuclear missiles to Cuba. I recall perfectly well the 

day when another adviser ... told me that we were contemplating the 

idea of deploying nuclear missiles in Cuba if the Cuban leadership 

agreed. I was definitely taken aback with this information, because 

being someone knowledgeable of U.S. affairs, and realizing the 

importance of such a step, I knew this would entail serious 

consequences. I was then faced with the dilemma of discussing this 

with Khrushchev or not, although my colleagues in the Secretariat 

told me that there was no sense in discussing this because the decision 

had been made and a change in the decision would be impossible. 

However, I was doing this just to calm my own conscience, and when 

I found the appropriate time I talked to Khrushchev. He said that he 

was aware that this was a very serious decision, but why should we 

not do what the Americans had been doing all along? Why couldn't 

we deploy missiles in Cuba when we were surrounded by US military 

bases and US missiles? He even said that the Americans appealed to 

the Monroe Doctrine every so often, but that the Monroe Doctrine did 

not just call for the non-intervention of European states in U.S. affairs, 

but also the non-intervention of the U.S. in European affairs. The 

United States had already discarded this doctrine. 3! Against this logic, 

what was I to say, especially since I really did not expect a change in 

the decision that had been made? That was the end of our discussion. 36 

What is interesting here is Khrushchev's academic, legalistic, and deductive 

approach to the problem. In effect, Khrushchev thought that if the 

Americans were reasonable, they would tolerate the Soviet deployment. 

The Americans were reasonable; ergo, they would tolerate the deployment. 

This was clearly unreasonable. Even if Khrushchev's minor premise were 

correct, the major premise evinced an astonishing naivete about American 

sensibilities. Troyanovsky, Soviet Ambassador to the United States Anatoly 

Dobrynin, Aleksandr Feklisov, and no doubt other intelligence and foreign 

service officers who devoted their lives to the American political scene, 

could have corrected Khrushchev's astounding misapprehension about 

American complaisance." But he asked none of them. 

Fursenko and Naftali explain the failure to task the KGB to estimate the 

US response in terms of the organizational culture of intelligence in the 

Soviet Union: 

For Western scholars, it is even more startling that no one in the KGB 

was asked for a formal assessment of the likely American reaction to 

the placement of missiles in Cuba. Since World War II, the US



intelligence community had attempted to provide estimates of the 

intentions and capabilities of important foreign countries. While these 

estimates often evinced the flaws of committee-generated reports, 

they at least reflected an effort to coordinate and make sense of secret 

knowledge on a certain policy matter. In some cases, these reports 

sought to clarify the murky world of the subjunctive, to illuminate 

what might happen. Called 'intelligence estimates' in the Anglo- 

Saxon intelligence world, these reports were foreign to Soviet 

experience. In Moscow, 'what if questions were reserved for the 

consideration of the Presidium alone. 38 

This does not strike us as quite right. 'What ifs' are an inherent part of threat 

assessment. As Fursenko and Naftali themselves note, the KGB was tasked 

to provide an assessment of US intentions toward Cuba several times - for 

example, immediately after Aleksei Adzhubei's meeting with President 

Kennedy in January 1962" - and in some of these assessments, the KGB 

attempted to identify hypothetical conditions under which the United States 

might attack. An especially interesting example of this was a July 1960 

estimate that a US invasion of Cuba was unlikely unless the Cubans invaded 

the US naval base at Guantanamo Bay or unless the Soviets attempted to 

station nuclear missiles in Cuba." 0 This estimate - which Fursenko and 

Naftali imply had never formally been withdrawn or revised — was, in our 

view, a quality piece of intelligence that certainly should have given 

Khrushchev pause had he not been so convinced (for whatever reason) that 

the deployment would succeed. 

While it may well have been unusual for Soviet leaders to pose direct 

'what if questions to the KGB, it was not unusual for the KGB to consider 

them. But more to the point, the organizational culture of Soviet intelligence 

cannot explain this particular failure well because the real question is not 

simply 'why was the KGB not tasked to assess the likely US response to a 

nuclear missile deployment?', but why was nobody asked? Quite simply, 

Khrushchev did not think it necessary. He was the kind of man - headstrong 

and brimming with self-confidence — who did not feel the need to ask the 

experts. 41 

2. Did intelligence play any role in Khrushchev's judgment that the 

deployment would succeed? Troyanovsky's testimony highlights 

Khrushchev's deductive approach to forecasting; but there was intelligence 

available that could have buttressed Khrushchev's confidence. A 'highly 

regarded source' indicated to Feklisov in March 1962 that Kennedy would 

not risk military action against Cuba before the mid-term congressional 

elections in November. 42 This report could have encouraged Khrushchev to



believe that he had an eight-month window of opportunity to act. Of course, 

as Fursenko and Naftali note, the KGB provided the Kremlin with ample 

material that could just as reasonably be construed to indicate American 

hostilities were likely at any time. Therefore, if Khrushchev had been 

paying attention to Soviet intelligence when attempting to gauge the likely 

American response, he was clearly paying attention selectively. 

3. Why was the GRU not tasked to assess the prospects for a successful 

covert deployment? Quite distinct from the question of whether the United 

States would tolerate Soviet nuclear missiles in Cuba was the question of 

whether the Soviets stood a decent chance of transporting them, building 

facilities for them, and making them operational before the United States 

could discover them. Khrushchev seems to have believed quite strongly that 

it was necessary to deploy covertly and to confront the United States with a 

fait accompli. Despite his academic and legalistic turn of mind, he 

energetically resisted when, in late September, Che Guevara and Emilio 

Aragones pleaded with him to go public with the deployment, stressing the 

sovereign right of the Soviet Union to supply, and of Cuba to accept, any 

armaments whatsoever that they deemed necessary for Cuba's defense. 

They proposed this so as to deprive the United States of a pretext for 

action. 43 Presumably, Khrushchev thought Americans were reasonable with 

guns pointed at their heads, but not otherwise. 

Soviet military intelligence would have been the most competent body 

to assess the feasibility of a secret deployment. There is no indication, 

however, that the GRU was asked to conduct a study and make a report. It 

would appear that Khrushchev was satisfied to allow Biryuzov to make that 

determination himself while in Cuba in late May and early June. Biryuzov 

appears to have concluded that MRBMs could easily be hidden in Cuba, in 

caves and mountains, or disguised as palm trees. 44 It is unclear what, if 

anything, Biryuzov thought about the feasibility of hiding the roads, 

buildings, and launch pads associated with the sites, or of preventing the 

United States from noticing the removal of already-deployed weapons from 

the Ukraine and their shipment 8,000 miles into the United States' 

backyard. The problem of maskirovka was simply left to the planners of 

Operation 'Anadyr' and treated as an operational detail. 45 

As with the first Soviet puzzle, the best explanation for the failure to task 

the GRU with a more intensive study may be Khrushchev's over-confidence 

and commitment. It would certainly appear that the Soviet leadership did 

not appreciate the difficulties of a secret deployment, which may reflect 

ignorance of logistical technicalities and/or American capabilities. 46 We do 

not know whether it would have been unusual for the Kremlin to ask 

military intelligence to conduct feasibility studies for secret military



deployments, so it is difficult to gauge an organizational-cultural 

explanation. It would be interesting to know in any case what the GRU 

would have concluded about the prospects for covert deployment if it had 

been given the opportunity to study the problem closely. 

4. What explains the sloppiness and amateurishness of KGB reporting in the 

Cuban missile crisis? Fursenko and Naftali give an account of Soviet 

intelligence during the Cuban missile crisis that is astonishing given the 

KGB's reputation for competence. Students of intelligence will know that 

laymen and outsiders consistently overrate the capabilities of intelligence 

services; but even when measured against a more realistic standard, the 

KGB appears to have performed extremely poorly in this episode. Fursenko 

and Naftali describe an organization obsessed with espionage, unable to 

integrate secret and open sources, either unwilling or unable to synthesize 

and interpret, ill-informed about Soviet policy, lacking well-placed 

informants, reduced to the ignominy of hanging around a parking lot at the 

crack of dawn to keep a journalist under observation, and - if one of the 

more plausible versions of the Scali-Fomin story is correct - undisciplined 

to boot. 

While we can certainly blame Khrushchev personally for certain 

demand-side intelligence failures, clearly he cannot be responsible for 

supply-side failures. We are inclined to credit Fursenko and Naftali's 

organizational culture explanation here. 'Professional prudence was the 

principal reason for the inadequacy of Soviet intelligence analysis', they 

write, combined with a 'limiting institutional ethos' emphasizing the 

collection of secret information. 47 This goes quite a long way toward 

explaining why the KGB fared so poorly. But in addition to this, it is worth 

noting that the KGB suffered from technical shortcomings that made its 

performance seem all the worse. The KGB's inability to fathom the mind 

of John F. Kennedy did not reflect institutional incapacity per se, since a 

different ethos, and a different set of procedures, could not have 

compensated for a lack of quality sources at or close to the White House. 

But unlike the CIA - which was equally unable to fathom the mind of 

Nikita Khrushchev, and which appears likewise to have had no good 

sources close to the Kremlin - the KGB could not fall back on impressive 

technical collection of intelligence on the adversary's capabilities in order 

to contribute to policy making, and to appear busy and productive to 

national leaders. 

Cuban Puzzles 

We have so little information on Cuban intelligence — there was essentially 

none, prior to Domingo Amuchastegui's essay - that it is difficult even to



speculate about possible explanations for certain obvious questions, let 

alone hazard tentative answers. Nevertheless, several are worth flagging. 

First, Amuchastegui emphasizes the stark disagreement between Soviet 

intelligence and the Direction General de Inteligencia (DGI) on the 

likelihood of an American attack on Cuba. Amuchastegui's discussion of 

the various possible explanations for that disagreement is both fascinating 

and instructive. For the moment, however, it is worth recalling 

Amuchastegui's claim that Cuban military intelligence - the Direction de 

Inteligencia Militar (DIM) - consistently echoed the Soviets' dire threat 

assessments. 48 What, then, explains the stark disagreement between DIM 

and DGI? 

Both organizations had high-level Soviet advisers, so it seems unlikely 

that the difference can be accounted for simply by a closer professional 

relationship between Soviet and Cuban personnel in one case than in the 

other. DIM fell under the Ministry of the Revolutionary Armed Forces 

(MINFAR) led by Raul Castro, however, and it is conceivable that Raul's 

pro-Soviet sympathies may have had some influence on DIM's willingness 

to accept Soviet assessments. Another possibility is that DIM's professional 

focus on military threat assessment led it to rely heavily on American 

military capabilities as an indicator of political intent. DGI appears to have 

based its assessment on a broader analysis of political context, and on its 

reading of the goals of Operation 'Mongoose' (certainly a more defensible 

basis, since military capabilities are not reliable indicators of political 

intent). 

A second puzzle is Fidel Castro's failure to inform his own intelligence 

services of the true nature of the Soviet deployment. It is not clear why 

Castro would wish to keep his own intelligence service in the dark, although 

one obvious possibility is that he wished to limit dissemination of this 

information to reduce the danger of leakage. Given the state of unrest in 

Cuba and the intense activity of counterrevolutionary groups, it would have 

been reasonable for Castro to fear that the CIA had informants within his 

intelligence services. Yet Cuban intelligence could not properly fulfill its 

primary mandate - warning of an American attack - without knowing what 

the Soviets in Cuba were up to that might provoke one. 

This leads directly to a third puzzle: namely, if Cuban intelligence did, 

in fact, catch on to the nuclear deployment, and did attempt to warn the 

Cuban leadership of the danger that this might provoke the United States, 

why did Pineiro refuse to pass that threat assessment on to Castro? 

Amuchastegui explains this in terms of Pineiro's 'reluctance to submit to the 

Cuban leadership assessments that would challenge or even contradict Fidel 

Castro's own thinking'. 49 It is certainly possible that Pineiro worried Castro 

might shoot the messenger. Yet if this is the case, why did Castro send Che



and Aragones to Moscow in late August to plead for public disclosure? 

Surely Castro himself must have been concerned that the Soviet deployment 

could precipitate an American attack. It is far from clear that on this score 

the judgment of Cuban intelligence would challenge or contradict Castro's 

thinking. Pineiro may have known that, of course, in which case he would 

have been wise not to pass on a report containing nothing Castro did not 

already know or believe. This would only irritate Castro, and in Cuban 

politics, that would be a mistake. 

Fourth - and, for moment, finally - what was the basis of Castro's 

confidence on 26 October that an American attack was imminent? 

Amuchastegui insists DGI did not provide Castro with any information to 

that effect. Certainly it is unusual for political leaders to warn intelligence 

analysts of an impending assault. 50 Amuchastegui is probably correct to 

infer that the information did not come from Soviet intelligence, since 

Castro issued precisely the same warning to Khrushchev. It is possible, as 

Amuchastegui speculates, that Castro drew this conclusion from his own 

reading of open sources, or, as Fursenko and Naftali suggest, from Brazilian 

President Joao Goulart, whose explicit warning of an imminent attack 

coincided with, and reinforced, an otherwise inconclusive report from 

Cuban intelligence in New York. 51 As far as we know, Castro himself has 

never identified a particular source for his judgment. 52 It thus remains a 

mystery. 

EVALUATING INTELLIGENCE PERFORMANCE 

The authors of our three empirical essays all make judgments about the 

performance of the respective intelligence bodies they examine (and, in at 

least one case, of the adversary's intelligence service as well). Were they to 

assign grades, we suspect that Garthoff would give the American 

intelligence community an A-; Fursenko and Naftali would give the Soviet 

intelligence community at best a C- (if not an 'Absent'); and Amuchastegui 

would give the Cuban intelligence community a solid B+ (and the CIA a D). 

The Stennis Committee and the President's Foreign Intelligence Advisory 

Board (PFLAB) both probably would have given the CIA a grade of B-; 53 

DCI McCone, in his response to the PFIAB report, a B+. 54 We know of no 

official Soviet intelligence post mortems, so we cannot determine how the 

Soviets assessed their own intelligence performance. Amuchastegui notes 

that there were major changes to Cuban intelligence after the crisis, 

indicating some dissatisfaction, but that many of these changes simply 

sought to reduce Cuba's reliance on Soviet and East European intelligence. 

It is natural to try to judge the performance of intelligence communities, 

yet it is very difficult to know what standards or benchmarks to employ. It



is unrealistic to expect intelligence to know everything, and to know it right 

away. Omniscience, omnipotence, and ubiquity are out of the question. 

Even an intelligence service operating at peak efficiency and unhampered 

by resource constraints will make mistakes and fail to anticipate important 

events every once in a while. In her essay, Beth Fischer argues that perfectly 

normal human psychology imposes a 'performance limit' on intelligence. 

Two decades ago Richard Betts eloquently and perceptively identified a 

number of other insurmountable obstacles to perfection, psychological and 

otherwise. 55 

How, then, do we know how well an intelligence community is 

performing? We cannot design meaningful tests of real-world performance, 

as we can (say) for student drivers. There are two reasons for this. First, we 

cannot manipulate the international environment. Certainly it would have 

been an intriguing idea for American officials to ask Nikita Khrushchev to 

deploy nuclear missiles in Cuba, so that they could test the CIA's 

forecasting skills; but there would have been no way to secure Khrushchev's 

compliance, and in any case, this would have been perverse. Second, since 

we cannot predict the future, we cannot know the 'right answers' before we 

administer a test. If we could, we would not need intelligence services. 

Occasionally it is possible to know in real time when an intelligence 

service misses something. From time to time crucial information of the kind 

it is the intelligence community's job to acquire reaches national leaders 

through entirely different channels. This does not, however, provide a basis 

for determining whether intelligence communities are performing well or 

poorly in general. An intelligence community that is performing extremely 

well in general will, unavoidably, miss certain things. This is logically 

implied by the notion of a performance limit. 

Retrospective Evaluation: Three Temptations 

Most judgments about the performance of intelligence arise from 

retrospective evaluations, such as official post mortems and academic 

studies. Retrospective evaluation has the crucial advantage of a fixed 

historical record against which to assess analysts' perceptions, beliefs, and 

inferences. There is always some irreducible element of uncertainty about 

what actually happened, of course, but this degree of uncertainty cannot 

compare with our uncertainty about what is happening now and what will 

happen in the future. Retrospection is the best we have. 

How accurately can we gauge intelligence performance in retrospect? 

We submit that the answer is 'not very', largely for straightforward 

epistemological reasons. However, we would like to suggest that 

retrospective evaluation - though imperfect - can be useful if we have a 

particular understanding in mind of how we can benefit from it. We suspect,



however, that unrealistic expectations about intelligence performance, 

coupled with understandable but misguided views about what retrospective 

evaluation is for, typically frustrate the enterprise. 

Three common temptations represent obstacles to fruitful retrospective 

evaluation: the temptation to focus on the specific and the spectacular (a 

form of selection bias); the temptation to privilege hindsight; and the 

temptation to try to evaluate performance in terms of a 'rate of success.' 

1. Selection bias. It is almost irresistible to focus on particular dramatic 

failures when assessing intelligence performance retrospectively. The 

enormous literature on surprise attacks provides a useful example. 

Admittedly, these are engrossing events. It may be doubted whether studies 

of intelligence successes could compete for our attention, because one of the 

chief things that makes a failure spectacular is that it has a dramatic 

consequence. A properly-exploited intelligence success may lead to the 

prevention of something spectacular. Moreover, a success does not provide 

the occasion for public angst about intelligence capabilities, and would not 

normally spawn public review or discussion. Our fascination with failures 

is entirely understandable. However, it is clear that by focusing our attention 

on failures, we may overestimate the weaknesses of an intelligence 

community - much as aviation disasters incline people to exaggerate the 

dangers of flying (uneventful flights do not make headlines). 

In principle, we could avoid this temptation by making an effort to study 

a representative sample of intelligence episodes - selecting not on the 

outcome, but randomly, or by period of time. But this is easier said than 

done. Successes leave less of a documentary trail than failures, and 

mundane episodes attract less attention than dramatic ones. There may 

simply not be enough information available to conduct a truly balanced 

study. The combination of inherent allure and available information means 

that spectacular intelligence failures have an almost irresistible gravitational 

pull. 

2. Privileging hindsight. Hindsight tends to color our judgments of how 

intelligence analysts should have interpreted information. This leads to 

unrealistic expectations about performance. 

Once we know what has happened, we can easily distinguish between 

meaningful and meaningless indicators: the signals stand out starkly from 

the noise. But to intelligence analysts attempting to anticipate events, those 

same signals may be barely perceptible, and quite understandably so. 56 After 

the Japanese attacked Pearl Harbor, it was easy to see that there was plenty 

of information available on the basis of which US intelligence might have 

anticipated the attack. It is tempting to take the further step and fault US



intelligence for not anticipating it. Hindsight provides such a powerful 

prism that the intellectually frail often cannot resist the conclusion that US 

intelligence must have known about Pearl Harbor in advance. Thus are 

conspiracy theories born. 57 

It is in the nature of Monday-morning quarterbacking to argue that if 

intelligence had done X rather than Y, they would have plucked the signals 

from the noise and correctly anticipated events. And it may well be true that 

if US intelligence had concentrated on certain targets rather than others, 

relied more heavily on certain methods of collection than others, handled 

information differently, made use of different working assumptions, etc., 

etc., they would have reached the conclusion that the Japanese were about 

to attack Pearl Harbor in time to do something about it. The problem is that 

we cannot know this, even with the benefit of hindsight. All choices have 

trade-offs and opportunity costs that are difficult to estimate. Choosing X 

might have induced other (possibly cascading) errors. 

Moreover, to go the extra step to conclude that intelligence should have 

done X rather than Y begs additional questions. Was X technically possible? 

Would intelligence have had good reasons for doing X at the time? Would 

those reasons have been better than the reasons they already had for doing 

Y? How do we know what reasons are good, what reasons are bad, and, of 

two good reasons, which is the better? 58 

3. The temptation to determine a 'rate of success'. In an enlightening and 

deservedly well-regarded essay written shortly after the Cuban missile 

crisis, Klaus Knorr attempted to understand why American intelligence 

failed to anticipate the Soviet deployment, and to put that failure in a context 

that would permit a meaningful evaluation of intelligence performance. 

Knorr quite properly acknowledged the inevitability of surprise. In a nowfamous 

passage, however, he stated that 'the practical problem is to improve 

the "batting average" - say, from .275 to .301 - rather than to do away 

altogether with surprise'. 59 We believe this to be misguided. 

It is a natural temptation to think of 'good' intelligence performance in 

terms of a high rate of success, and 'poor' intelligence performance in terms 

of a low one. Just as we can be confident that a baseball player batting .350 

must be doing things right and a player batting .125 needs serious work, an 

intelligence agency that fails rarely must be doing well, and one that fails 

often must be doing badly. It is also natural to think that retrospective 

evaluation, despite its various pitfalls and difficulties, at the very least 

permits us to gauge the relative frequency of correct and incorrect 

judgments. On first glance, this seems straightforward. When an 

intelligence community correctly anticipates the enemy's move, we score a 

hit; when it does not - as the CIA did not, in failing to foresee the Soviet



nuclear deployment to Cuba - we score an out. 

The moment we begin to think about operationalizing these assessments, 

however, we quickly discover that the metaphor is both meaningless and 

unworkable. Consider SNIE 85-3-62: 

The USSR could derive considerable military advantage from the 

establishment of Soviet medium and intermediate range ballistic 

missiles in Cuba, or from the establishment of a Soviet submarine 

base there. As between these two, the establishment of a submarine 

base would be more likely. Either development, however, would be 

incompatible with Soviet practice to date and with Soviet policy as we 

presently estimate it. It would indicate a far greater willingness to 

increase the level of risk in US-Soviet relations than the USSR has 

displayed thus far, and consequently would have important policy 

implications with respect to other areas and other problems. 60 

The estimate equivocated. In effect, it said that a nuclear deployment was 

unlikely, but not impossible. It did not assign probabilities, but, for the sake 

of argument, we can imagine that SNIE 85-3-62 meant to communicate that 

there was a 0.2 probability that the USSR would deploy MRBMs to Cuba, 

and a 0.8 probability that it would not. When calculating the CIA's batting 

average, should we count this as one-fifth of a hit, and four-fifths of an out? 

DCI McCone estimated the reverse. Was he at bat, too, or only the authors 

of the SNIE? 

Suppose that Khrushchev had consulted with his American specialists 

earlier in the game, got cold feet, and decided to subtract the nuclear 

component from the Soviet deployment. In that case, SNIE 85-3-62 would 

have been right on the money, and we would have to score it a hit (or, at 

least, 80 per cent of a hit). And yet, in this scenario, there is nothing 

whatsoever that the CIA would have done differently. Do we want our 

evaluations of the performance of an intelligence community to turn on 

extraneous stochastic contingencies? Put another way: why should we not 

blame Khrushchev, rather than the Office of National Estimates, for the 

mistaken SNIE? 61 

Consider now Cuban intelligence. The DGI predicted that the United 

States would not tolerate the presence of Soviet nuclear missiles in Cuba. 

By 'not tolerate', they had in mind a military response. Now, the DGI was 

correct to predict that the United States would not tolerate the missiles, but 

it did not anticipate the blockade. How do we score this? Do we count it as 

two at-bats - one hit and one out for a batting average of .500? Or was the 

second estimate simply part of the first - a swing-and-a-miss during the one 

at-bat of consequence, so to speak - for an average of 1.000? 62 

Consider also that from April 1961 to September 1962, DGI estimated



that an American attack was not imminent. 63 During that period, no such 

attack took place. How do we score this? Do we count one successful at-bat 

for every day that the United States did not invade? Every week? Every 

month? Every three months? Our decision will certainly have dramatic 

implications for DGI's batting average. 64 

It is clear that there is no non-arbitrary, rigorous way to gauge the 

relative frequency of successful and unsuccessful judgments, even in 

retrospect. Even if there were, we could not be certain that this would be a 

valid indicator of the competence of an intelligence community. Sloppy 

estimates are sometimes right, and careful estimates sometimes wrong, for 

reasons over which the estimators have no control. The batting average 

metaphor is therefore misplaced. 

Indeed, it is possible to argue that a batting average is an imperfect 

indicator of skill even in baseball. Batters who play more games on artificial 

turf, against weaker pitchers, and against weaker defensive teams will have 

higher batting averages than better batters who play on grass against 

stronger pitchers and better defensive teams. Over a long season, a batting 

average can be a fairly reliable indicator of skill for players who play in a 

league with a balanced schedule, because these circumstances approximate 

the ceteris paribus condition. But even then a batting average is only a 

relative measure of skill, and the valid comparison group is limited to other 

players in the league. Intelligence analysis takes place under comparatively 

few parametric constraints. Analysts may be called upon to predict 

anything, anywhere, by anyone. Different intelligence communities may 

face radically different challenges - as James Wirtz discusses in his essay - 

and one intelligence community may face consistently more difficult tasks 

than another. Finally, in intelligence assessment there is no 'standard 

equipment', and corked bats, far from being forbidden, are actively 

encouraged. Intelligence 'batting averages', therefore, would be 

meaningless measures of relative performance even if it were possible to 

calculate them in a rigorous, non-arbitrary way. 

Knorr might reply that he did not intend for us to take him so literally. 

Perhaps he simply meant his metaphor to imply that the practical problem 

is to minimize the number of errors. Some might be tempted to say that this 

goes without saying. We are not so sure. It may not be the number of errors 

that best measures intelligence performance, but the kind of errors. 

Perils Illustrated: the US 'Failure' to Predict the Soviet Deployment 

It might be helpful at this point to examine an important judgment that has 

attracted special attention in the preceding essays (and elsewhere), to see 

some of these temptations at work, to illustrate the difficulties of avoiding 

them, and to highlight the difficulties of making rigorous evaluations. This



will help prepare the ground for our discussion of how retrospective 

evaluation can be used constructively. Because of the richness of the 

available record, the US 'failure' to predict the Soviet deployment will 

serve well. 

The Stennis Report concluded that the 'predisposition of the [US] 

intelligence community to the philosophical conviction that it would be 

incompatible with Soviet policy to introduce strategic missiles into Cuba' 

was the primary cause of its failure to predict the deployment. 65 This 

preconception, according to the report, prevented the intelligence 

community from giving 'proper weight' to 'indications to the contrary'. 66 

The implication is that US intelligence should not have been so firmly 

wedded to its preconceptions, and if it had not been, it might not have made 

this important error. 

The Stennis Report's conclusions provide an interesting example of 

Monday-morning quarterbacking. Did the US intelligence community have 

'good reasons' for relying less heavily upon the preconception that such a 

deployment would be incompatible with Soviet strategic policy? Garthoff 

writes: 'In the absence of clear evidence of deployment, the estimate [i.e., 

SNIE 85-3-62] had to rest on an assessment of Soviet intentions, and the 

past record tended to support the conclusion that the Soviet leaders would 

not deploy strategic nuclear missiles in Cuba.' 67 Our sense is that most 

professional intelligence analysts, and most scholars of the Cuban missile 

crisis, would concur with this judgment. 

Garthoff's statement has several interesting features. The first is the 

phrase, 'In the absence of clear evidence of deployment'. What, exactly, is 

'clear evidence of deployment'? 'Only direct evidence, such as aerial 

photographs, could be convincing', Garthoff writes. 68 As a psychological 

matter, this was undoubtedly true in 1962. Neither President Kennedy, nor 

senior members of the intelligence community - DCI McCone included, we 

suspect, since he was so concerned about stepping up photographic 

coverage of the island - would have been willing to conclude that the 

Soviets were deploying missiles to Cuba without photographs of Soviet 

missiles in Cuba. But ought American officials to have had such a 

demanding standard of proof? The Stennis Report did not state, but seemed 

to imply, that the answer was no, concluding that a contributing cause of the 

failure was the CIA's tendency 'to discredit and downgrade the reports of 

Cuban refugees and exiles'. 69 As we have already suggested, if the CIA 

discounted reports simply on the ground that they came from Cuban 

sources, this would at least have been a logical error, and might also have 

been a pathological prejudice. 70 But would it have been an avoidable one? It 

is easy to say in retrospect that the CIA should not have been so hard on 

Cuban refugee reports, but it is difficult to imagine how CIA analysts



working without the benefit of hindsight could have 'corrected' for a 

prejudice of which they were not conscious, and which, though illogical, is 

perfectly normal psychologically. 71 

The second interesting feature of Garthoff's rejoinder is the phrase, 'the 

estimate had to rest on an assessment of Soviet intentions'. If the CIA could 

not see that a deployment was underway, it would have to try to read Soviet 

minds. The CIA had no sources in the Kremlin, and hence no good 

information on whether Khrushchev and his colleagues were contemplating 

a nuclear deployment. How, then, were they to know what Khrushchev was 

thinking? 

Enter the third interesting feature of Garthoff's statement: 'the past 

record tended to support the conclusion that the Soviet leaders would not 

deploy strategic nuclear missiles in Cuba'. Garthoff here invokes the 

principle of induction: namely, that the future will resemble the past. 

Several logical difficulties arise here. First, the CIA was attempting to 

ascertain the likelihood of something that had never happened before, and 

to do so relied upon a principle whose strict application will always result 

in failure to anticipate the unprecedented. Second, the CIA made the logical 

mistake of inferring intentions from behavior. It is fallacious to infer present 

intentions from present behavior, let alone to infer present intentions from 

past behavior. An act, a practice, or a policy may be consistent with a variety 

of incompatible purposes (the deployment itself was logically consistent 

with both offensive and defensive motives, for example). 72 Third, even if 

past behavior could predict present intentions and future behavior, the 

inference would be arbitrary unless we could determine which of several 

possible ways of setting up the induction applied in a given case. It is 

certainly correct, as the CIA noted and emphasized, that the Soviet Union 

had never before deployed nuclear missiles beyond its borders. But neither 

had the Soviet Union ever failed to extend its nuclear umbrella over a 

friendly Marxist-Leninist state. Nor had the USSR ever passed up an 

opportunity to deploy land-based nuclear missiles in the Western 

Hemisphere (Cuba was, after all, the first opportunity to arise). Finally, 'past 

Soviet behavior' was a composite of the behavior of several Soviet leaders, 

only one of whom was Khrushchev, and he had not yet had much of an 

opportunity to leave his imprint. As Beth Fischer notes, the CIA might have 

attempted to base the estimate specifically on an assessment of 

Khrushchev's intentions and Khrushchev's risk-taking propensities, instead 

of on some generic 'Soviet' characterization. 73 There was no guarantee that 

Khrushchev would do what Lenin or Stalin would have done had they been 

in his shoes instead. 

Most US officials were shocked by the discovery of the deployment 

because it seemed to them to be so out of character. Nevertheless, it was



trivially easy for analysts working with the benefit of hindsight to construct 

explanations of Khrushchev's behavior that conformed with their 

preconceptions. Some of these retrospective assessments cut the Gordian 

knot simply by assigning blame to Khrushchev for failing to appreciate the 

nature and intensity of American concerns about Soviet forces in Cuba, or 

for overestimating the likelihood that a covert deployment might succeed. 

The current record shows that Khrushchev erred, but it is interesting to note 

that these retrospective assessments - while in this respect quite accurate - 

themselves made various indefensible logical leaps. Knorr, for example, 

making use of a constructive distinction between technical and behavioral 

surprise (that is, surprise that results because someone succeeds in fooling 

us, and surprise that succeeds because it is, or seems to be, inconsistent with 

our set of expectations), 74 concludes that there was nothing wrong with the 

set of expectations about Soviet behavior prevalent in the US intelligence 

community in 1962, and that the Stennis Report erred in finding fault with 

it. Instead, Knorr considers the failure to anticipate the deployment a case 

of 'apparent behavioral surprise', where the USSR seemed to have adopted 

a course of behavior that was inconsistent with US expectations, but, in fact, 

was not. In support of this conclusion, Knorr reasons that the lengthy period 

of planning and preparation that must have been involved in the deployment 

was 'not conducive to a strong component of emotional or irrational 

decision-making', and that Moscow's underestimation of the risk of the 

deployment can most probably be attributed to Soviet intelligence failure. 75 

Knorr is undoubtedly correct to suppose that there were Soviet intelligence 

failures; but we now know that the Soviet mistake was over-determined, 

and that emotional or 'irrational' factors most certainly played a role. 

Other retrospective assessments downplayed Soviet errors and quite 

creatively constructed a set of beliefs on the basis of which Khrushchev 

'must' have been operating that would have made the covert deployment 

appear to be a rational gamble. These are interesting because they managed 

to preserve three key assumptions: (1) that Khrushchev was aware of the 

nature and intensity of the American interest in Cuba; (2) that Khrushchev 

would not behave fundamentally differently from his predecessors; and (3) 

that Khrushchev would not accept a poor gamble. These explanations 

typically relied heavily (for example) on the claim that Khrushchev had a 

low estimate of Kennedy's resolve - because of his youth, his 'liberalism', 

his failure to follow up the Bay of Pigs disaster with full-scale American 

intervention, his overly-intellectual performance at the 1961 Vienna 

summit, and so on - and/or the claim that Khrushchev thought that his 

stranglehold on Berlin gave him the leverage he needed to force Kennedy 

to accept a nuclear missile deployment in Cuba. These assessments also 

make enormous logical leaps, egregiously demonstrating the fallacy of



inferring motives and calculations from behavior. 76 Even nuanced 

retrospective analyses that made some allowance for Soviet miscalculations 

tended to make indefensible assumptions and inferences that we now know, 

on the basis of a much richer record, to have been in error. Most strikingly, 

as Garthoff discusses at length, US post mortems consistently neglected or 

downplayed the extent to which Khrushchev's behavior was consistent with 

defensive or deterrent purposes. 77 It is interesting to note that every time US 

analysts raised the possibility - in real time or in retrospect - that 

Khrushchev might be operating on the basis of defensive motivations, they 

let this possible explanation wither on the vine. 

The conclusion we must draw, then, is that there was no logically valid 

way for the CIA to infer Khrushchev's intentions from the information 

available. 78 The information was consistent with any number of possible 

explanations and predictions. The best the CIA could do, as a matter of 

logic, was to say that Khrushchev might deploy nuclear missiles to Cuba, 

but that he might not. They could also say, in effect, 'If we were in 

Khrushchev's position, knowing what we know, we would not attempt to 

deploy nuclear missiles to Cuba.' We think this is not a bad way of 

interpreting SNIE 85-3-62. The difficulty is that it tells us a great deal about 

the CIA, but very little about Khrushchev. 

And yet it is difficult to specify exactly what the CIA should have done 

differently, given existing constraints on information. When the President of 

the United States asked whether the Soviet Union was likely to station 

nuclear missiles in Cuba, the CIA had a duty to try to answer. To predict the 

deployment correctly, the CIA would have had to assume a number of 

cockamamie things - for example, that Khrushchev would not ask any of 

his experts about how the United States would react to a nuclear deployment 

in Cuba, and that Khrushchev could believe that the Soviet military could 

send 45,000 men and dozens of ballistic missiles 8,000 miles into the 

Caribbean without the United States catching on. It is easy to imagine how 

the President of the United States would have reacted to an estimate that 

stated these two assumptions explicitly. 

Even if the CIA had fully factored Khrushchev's risk-acceptance and his 

defensive or deterrent motivations into its estimates, it is not at all clear that 

the CIA would have concluded that such a deployment was, on balance, 

probable. It was a breathtaking gamble. We are inclined to agree with 

Garthoff that, even taking into account what was not then known, the 

estimate might have concluded that Soviet leaders probably would not place 

nuclear missiles in Cuba, even while giving 'more weight and attention to 

the possibility that they would do so'. 79 But it is unlikely that this would 

have made any practical difference to a President already keenly intent on 

monitoring the Soviet buildup in Cuba. 80



Failures Consequential and Otherwise: How Should we Evaluate 

Intelligence Performance? 

It is ironic that the US intelligence community made such an intense effort 

to look for something that they did not expect to see. And while it is possible 

that the United States might have discovered the Soviet nuclear deployment, 

and detected the true size and nature of the Soviet conventional deployment, 

somewhat earlier than it did, had it relaxed its expectations, its standards of 

proof, or both, it is worth recalling - as Garthoff puts it - that 'intelligence 

did identify the missiles in sufficient time to allow successful American 

initiative and action to compel their withdrawal'. 81 'Discovery a week or 

two earlier in October,' Garthoff writes, '... would not have changed the 

situation faced by the President and his advisers'. 82 

In what sense, then, did the CIA 'fail'? It certainly failed to make a clear 

prediction that the Soviet Union would attempt to deploy nuclear missiles 

to Cuba. But it did not fail American policy makers who deeply dreaded the 

prospect. The CIA was vigilant, and told the President all that he really 

needed to know. Certain things the CIA did not know, and could not tell the 

President, even though he would have preferred knowing them: for 

example, exactly when the Soviet MRBMs in Cuba would become 

operational; whether or not nuclear warheads had yet arrived in Cuba, and 

if so, where they were stored, and whether they had been issued to missile 

units; and exactly how many Soviet troops were on the ground. We can call 

these failures if we like; but we are not certain that we see the point of doing 

so. Only the last was something the CIA could have been, able to divine 

given its existing capabilities, and given the information at its disposal. The 

others are very difficult tasks, and the CIA could only have known these had 

it been lucky in data collection. 83 But in any case, none had a negative policy 

consequence. As Garthoff notes, the White House, the Pentagon, and the 

CIA all assumed in prudence that the Soviet missiles were operational and 

that warheads were present, and given the nature of US contingency plans 

for military operations against Cuba, it is unlikely that the underestimate of 

the number of Soviet soldiers there would have materially affected the 

outcome of an invasion. 84 

The only significant lacuna that we can identify concerns the presence 

of tactical nuclear weapons. American officials did not presume that Soviet 

forces in Cuba would be equipped with tactical nuclear weapons, and had 

Kennedy sent American forces into Cuba, they would not have been 

equipped with similar weapons themselves. But while this was certainly a 

lacuna, it is unclear that the CIA should have been expected to discover that 

the Soviet Union had equipped its forces with tactical nuclear weapons, and 

it is possible to argue that the CIA had furnished Kennedy with ample



information on the basis of which to conclude that an invasion carried with 

it unacceptable risks in any case. Since he did not choose to invade Cuba, 

the CIA's failure to establish the presence of tactical nuclear weapons did 

not have negative consequences. In fact, as Garthoff argues (correctly, in 

our view), the CIA's failure to establish the full size and character of the 

Soviet deployment had a distinctly positive policy consequence: it 

facilitated the resolution of the crisis. It may be an unusual circumstance 

when intelligence serves policy better by knowing less rather than more; but 

this is a useful reminder nonetheless that the task of intelligence is not to 

know everything, but to serve policy, and that intelligence can serve policy 

well even when it does not know everything, and sometimes even when it 

makes mistakes. 

What we see on the American side of the Cuban missile crisis, at least, 

is a constructive intelligence-policy dynamic. Communications between the 

White House and the CIA were far from perfect, and there was a great deal 

of uncertainty in the air. But while the CIA did not predict a Soviet nuclear 

deployment, and did not fully appreciate the size, purpose, and battlefield 

capability of the Soviet expeditionary force even after identifying MRBMs 

in Cuba, it managed to avoid communicating any overconfidence in what it 

did know (or believe) that would have either encouraged recklessness, or 

undercut the President's respect for, and appreciation of, the intelligence 

process. Taking in the broad sweep of the CIA's assessments before, during, 

and after the acute phase of the crisis, we believe that it did quite an 

impressive job of serving US policy, in the light of its technical capabilities 

and the inherent limits of intelligence assessment. We can try to learn from 

its failure to predict the Soviet nuclear deployment, but it should not bear 

more than marginally on our evaluation of its performance over the course 

of the crisis as a whole. 

What of Soviet and Cuban intelligence? There is a sense in which Soviet 

intelligence was never really given a chance to fail, at least prior to the acute 

phase. Intelligence cannot serve policy when policy makers do not take it 

into their confidence and ask it policy-relevant questions. Once the crisis 

broke, it would appear that Soviet intelligence strove to give the Kremlin 

information useful for managing the crisis, but came up with very little. 

Certainly the KGB and GRU cannot be faulted for not having technical 

capabilities to rival CIA's or NSA's. But given the Soviet investment in 

Humint, one might have expected a better performance than Fursenko and 

Naftali describe. We will never know, of course, what contribution to policy 

Soviet intelligence might have made had it better integrated and synthesized 

sources, and had it not engaged in self-censorship. But we do know that 

these shortcomings deprived it of any real possibility. Our expectations of 

the performance of Soviet intelligence must be somewhat lower than for



American intelligence to begin with, in view of its technical inferiority and 

the acute challenges it faced by virtue of the structural and domestic 

political circumstances under which it operated (as Wirtz discusses at 

length). Nevertheless, even when measured against a lower standard of 

reasonable expectations, Soviet intelligence hardly acquitted itself well. We 

may say in its defense, however, that while it had great difficulty being 

relevant, at least it cannot be held responsible for disastrous policy. 

We cannot hold Cuban intelligence up to a standard of performance 

appropriate either to American or Soviet intelligence. It was a nascent 

community operating under severe technical and resource constraints. 

Moreover, the Cuban leadership was clearly selective in what questions it 

asked its intelligence community, how it listened, and how it used 

intelligence. Amuchastegui's account suggests that most of the pathologies 

evident in the Cuban intelligence-policy relationship lay on the policy side. 

Nevertheless, as Amuchastegui makes clear, Cuban intelligence did make 

mistakes. They were not, however, mistakes of the kind that would 

encourage faulty policy, and Amuchastegui's tale indicates to us that Cuban 

intelligence had a remarkably mature demeanor throughout the episode in 

question. It was circumspect, relatively open-minded, and acutely aware of 

the structural and domestic political challenges it faced. In professionalism, 

it certainly compared favorably with the KGB. 

Evaluating the performance of an intelligence community in this way - 

rather than focusing intently on specific spectacular judgments (usually 

misjudgments), second-guessing analysts with the benefit of hindsight, or 

attempting to gauge a rate of success - has, we believe, at least four merits. 

First, it forces us to factor into our evaluations some baseline expectation of 

performance. Clearly intelligence communities differ from each other in 

raw capability and the challenges they face. Simply counting successes and 

failures - even if we could do it meaningfully - could tell us nothing about 

how well they are doing relative to each other. But at least we can determine 

roughly how well an intelligence community is doing relative to its own 

performance in the past when we begin by trying to establish a reasonable 

expectation. Second, while judgments of this kind can never be scientific, 

and while judgments of what is a 'reasonable' expectation are bound to be 

somewhat impressionistic, this style of evaluation has the virtue of 

highlighting the notion of a performance limit. All intelligence communities 

are going to make mistakes. Third, by forcing us to think about performance 

over an extended period of time, rather than with respect to a specific 

judgment, it makes it easier to bear in mind that at the end of the day the 

crucial judgment we wish to make is how well intelligence serves policy, 

not how well intelligence performs in purely intellectual exercises such as 

forecasting. Fourth, by evaluating performance over an extended period of



time, it is possible to discern habits and tendencies to cultivate or to try to 

eliminate, strengths to exploit in the future, and weaknesses to try to shore 

up. 85 No clear practical implication follows from the conclusion that the 

CIA failed to predict the Soviet nuclear deployment. Clear practical 

implications follow from the conclusion that Cuban intelligence depended 

too heavily on Soviet and East European technical services, that the KGB 

relied too heavily on espionage, and that the CIA was excellent at technical 

monitoring but weaker at political analysis. We learn these things only by 

taking in the broad view. 

Monday-morning quarterbacking and attempting to improve the batting 

average, then, are misguided approaches to evaluating intelligence 

performance. We offer instead a third sporting metaphor drawn from the 

only game harder than life itself- golf. As a metaphor, it has a few obvious 

weaknesses, but we believe that it also has much to commend it. So as not 

to try the non-sporting reader's patience unduly, we will simply highlight a 

few relevant points about the game of golf, leaving readers to tease out for 

themselves the isomorphisms: 

• The best golfers and the worst golfers all play the same courses. There 

is a sense, then, in which their scores are directly comparable to one 

another. Moreover, built into the very nature of golf is a performance 

limit. Excellent golfers will consistently beat par, but no golfer can even 

in principle score lower than 18 in one round of golf, and certainly no 

golfer playing a standard par-72 course with regulation equipment could 

ever hope to score (say) 36. Indeed, scores in the low 60s are rare. 

• While all golfers' scores are directly comparable to one another, the 

overwhelming majority of golfers compete not against others, but 

against themselves. A 'good' round of golf is one in which one scores 

lower than usual; a 'poor' round is one in which one scores higher than 

usual. (Amateur golfers who do compete against each other sometimes 

factor in their handicaps - constrained averages of their previous scores 

- so that, no matter what the skill differential may be, they can make a 

game of it. Notice that even in this case, they are playing against 

themselves at the same time as they play against their opponents, 

because the player who scores best relative to his or her own average 

round wins.) 

• A golfer can make what for him or her would be an excellent score 

despite making some number of poor shots. Indeed, a round without a 

single poor shot is almost unheard of, even for professionals. The best 

golfers are those who make relatively few poor shots, and recover from 

them well.



• The game has an important strategic element in addition to its simple 

mechanics. Players with beautiful swings who consistently choose the 

wrong clubs or make poor decisions about where to aim - players, for 

example, who refuse to acknowledge the existence of tall trees and water 

hazards - may score worse than players whose swings are technically 

inferior, but who play a smarter game. 

• Golfers have different styles, and different strengths and weaknesses. 

Some are risk-averse, and some are risk-acceptant. Some are better off 

the tee; others are better around the green; still others are better on the 

green. Some have no difficulty with sand traps and do not take pains to 

avoid them; others have great difficulty and try to avoid them at all costs. 

• Golfers rarely play only one course, though typically they become quite 

familiar with some number of courses. They enjoy the challenge of 

playing new courses, though they understand that, given two courses of 

approximately equal difficulty, they cannot expect to score as well on the 

course that they have never played before. 

• Golfers are strong believers in post mortems. When they score well, they 

like to try to identify and explain particularly good shots (often these are 

recovery shots). When they score poorly, they reflect solemnly upon 

their poor shots - though they may not choose to do so publicly. Golfers 

try to cultivate practices that work well, and to eliminate bad habits. 

• All golfers profoundly appreciate the role of luck. 

CONCLUSION: THE 'THEORY' AND PRACTICE OF INTELLIGENCE 

ASSESSMENT 

MENO: How will you look for it, Socrates, when you do not know at 

all what it is? How will you aim to search for something you do not 

know at all? If you should meet with it, how will you know that this is 

the thing that you did not know?" 6 

In 1978, Richard Betts lamented the lack of a 'theory' of intelligence. Betts 

suggested that '[njegative or descriptive theory - the empirical 

understanding of how intelligence systems make mistakes - is well 

developed,' but that positive theory and normative theory were not. 'The 

distinction is significant', Betts wrote, 'because there is little evidence that 

either scholars or practitioners have succeeded in translating such 

knowledge into reforms that measurably reduce failure. Development of a 

normative theory of intelligence has been inhibited because the lessons of 

hindsight do not guarantee improvement in foresight, and hypothetical



solutions to failure only occasionally produce improvement in practice.' 87 

We might quibble with the notion that failure can be reduced 

'measurably', but otherwise Betts's point is well-taken, and we see no 

evidence that scholars have managed to redress the lack of theory that he 

lamented. Certainly if by 'positive theory' we mean a robust and heuristic 

set of propositions relating causes and effects - e.g., condition x predicts y 

with probability p - it is impossible to identify anything in the open 

literature that would count as a 'theory' of intelligence assessment. And we 

may justly wonder whether any such thing is possible in principle. The very 

difficulty of operationalizing the notion of 'success' and 'failure' would 

seem an insurmountable obstacle. But this is the least of it. Intelligence 

assessment is not, strictly speaking, a science. It is an art. Intelligence 

analysts are not in the business of drawing logically-necessary conclusions 

about behavior on the basis of evidence generally knowable to be sufficient 

for that task. 88 Leaders of states do not make decisions under constraints 

sufficiently numerous and sufficiently knowable to make talk of the 

'probabilities' of their choosing alternative courses of action objectively 

meaningful. 89 Such language conveys information not about the target, but 

about the analyst. 90 When McCone and his subordinates at the Office of 

National Estimates debated whether Khrushchev was 'likely' or 'unlikely' 

to deploy MRBMs to Cuba, they articulated differences in their own fears 

and beliefs, and their own judgments of the implications of what American 

intelligence saw on the high seas and in Cuba. Khrushchev, however, was 

not literally rolling dice. There was no objectively knowable 'likelihood' 

that he would do anything in particular. McCone was right for bad reasons, 

and his analysts were wrong for less-bad reasons, but Khrushchev was 

Khrushchev - a complicated man with a mind of his own, not some 

automaton preprogrammed according to a 'Soviet' behavioral algorithm. 

Betts is clearly correct that '[n]egative or descriptive theory - the 

empirical understanding of how intelligence systems make mistakes - is 

well developed', if by this he means that there are many different bodies of 

theory upon which to draw in order to try to explain errors. Fischer's essay 

provides a good example. The style of argument whereby one assesses the 

'goodness of fit' of a body of propositions well-established in one domain 

with the empirical details of a particular case in another does not, however, 

lead to the kind of cumulation or generalization to which positive science 

aspires. Nor does the application of any one body of theory to a series of 

cases of intelligence failures tend to satisfy (e.g.) a Lakatosian criterion of 

theoretical 'progress'." It is often possible to say that a body of theory can 

help us make sense of an otherwise puzzling event, because it supplies us 

with concepts and cause-effect propositions in terms of which to understand 

it. But the difficulty is that there may be many different bodies of theory that



can do this, or that the body of theory that does the best job of helping us 

make sense of one puzzle may do a poor job, vis-a-vis some other body of 

theory, with respect to another. 

Without a positive theory of intelligence assessment, it is difficult to 

know how to go about constructing normative theory. Put another way, there 

can be no engineering without natural science. This is why it may be 

unfortunate that 'official post mortems of intelligence blunders inevitably 

produce recommendations for reorganization and changes in operating 

norms'. 92 Not only can we not be certain that we would have avoided one 

mistake had the organization and process of intelligence been different in 

the particular case at hand, but our tinkering may well induce other, possibly 

more serious mistakes in the future. 

What then, can we know in general about intelligence assessment, and 

how can this knowledge help improve the performance of intelligence? 

Recognizing that there is a performance limit to intelligence, and that 

intelligence assessment is an art rather than a science, should certainly 

induce caution against lofty expectations. But there is one general 

phenomenon characteristic of human judgment and perception the greater 

appreciation of which can cultivate a useful circumspection in intelligence 

assessment. Before closing, we will touch upon it and reflect upon how it 

might bear on practice. 

When Meno asked Socrates, 'How will you look for it ... when you do 

not know at all what it is?', he did not make the mere 'debater's argument' 

for which Socrates chided him. 93 Meno asked a profound question: how can 

we have knowledge of something without some prior familiarity with it? 

Once we allow some prior familiarity, we court infinite regress and the 

disturbing conclusion that knowledge is not possible at all. Socrates's 

solution — that all learning is mere recollection, for the immortal soul knows 

everything 94 - would seem on the face of it to substitute one paradox (the 

paradox of time) for another (the paradox of knowledge). In any case, the 

metaphysical presuppositions of Socrates's answer seem in no less want of 

justification than the answer itself. 

The paradox remains unresolved, though its contours and implications 

are now clearer, thanks to the efforts .of Gadamer, Quine, Chomsky, and 

others. 95 For purposes of understanding intelligence assessment, the paradox 

is most enlightening when we throw it into reverse: we do hold certain 

beliefs (sometimes we like to call this 'knowledge'), and they do influence 

how we understand new things. There are many vocabularies to use when 

describing and exploring the phenomenon and many angles from which to 

look at it. Fischer describes it in the language of cognitive psychology, but 

we could just as well do so in the language of structural sociology, 

semiotics, cultural anthropology, or hermeneutics. 96



The insight that our own concepts and beliefs affect our interpretation of 

the behavior of others can incline intelligence analysts to reflect on the 

perspective they bring to bear on an intelligence problem, and the 

possibility that, in the absence of very good quality evidence about others' 

deliberations and motives, their judgments may say more about themselves 

than about their targets. This does not necessarily mean that analysts will 

have an easy time trying to achieve some 'perspective on perspective', 

however. Garthoff writes: 

Today, I believe it is clear that there was a failure of estimative 

empathy, of viewing the situation as the Soviet leaders did as they 

made their decision to place missiles in Cuba. In particular, even if 

Soviet considerations of redressing the unfavorable strategic global 

balance for defensive and offensive purposes was paramount, as I 

continue to believe it was, it is now evident that deterrence of a US 

attack on Cuba was also a consideration of major if not equal 

importance. Yet the failure to see the situation through Soviet (to say 

nothing of Cuban) eyes blinded the US intelligence community as a 

whole to that significant factor. American intelligence estimators 

should have recognized a Soviet incentive to defend Cuba against a 

plausible American threat well before the Soviet decision to deploy 

missiles was made. 97 

Certainly if American decision-makers had appreciated that Khrushchev felt 

boxed in, they might have considered an attempt to deploy nuclear missiles to 

Cuba rather more likely. As Garthoff supposes, this might have given the 

President additional reasons to try to forestall it by means of earlier and 

stronger warnings. 98 But how should we interpret Garthoff's claim that the 

CIA 'should have recognized a Soviet incentive to defend Cuba against a 

plausible American threat well before the Soviet decision to deploy missiles 

was made'? As a retrospective lament, we endorse it completely. If Garthoff 

means to suggest that this was an avoidable mistake, however, we are not so 

sure. There were many reasons why US intelligence did not and could not 

easily empathize with Khrushchev. It took the shock of the Cuban missile 

crisis to demonstrate to American officials that they badly misunderstood 

him, and even then they did not appreciate exactly how they had 

misunderstood him. 99 It is difficult enough to empathize with an adversary in 

retrospect; it is asking a great deal of analysts to empathize in prospect. 100 

The paradox of knowledge must incline us to disagree with Fischer that 

there is a useful distinction to make between theory-driven and data-driven 

thinking. There is no such thing as data-driven thinking. Facts never speak 

for themselves. We can only make sense of the world on the basis of some 

prior understanding. There is a distinction to make, however, between



people who are tightly wedded to their beliefs, and people who are 

relatively more open to the possibility that their beliefs and assumptions 

may be wrong, or who appreciate the potential explanatory power of two or 

more different sets of beliefs and assumptions. 101 

Intelligence analysis can never transcend its dependence upon prior 

assumptions and beliefs. Understanding presupposes them. Yet at the same 

time, understanding must always, to some extent, be frustrated by them. All 

intelligence communities would like to minimize the frustration, and to tip 

the balance between knowledge of self and knowledge of other in the 

direction of the latter. As a practical matter, though, how do intelligence 

analysts avoid tripping over their prior assumptions and beliefs without 

being paralyzed by ambiguity and indecision? How do they steer between 

the Scylla of overcommitment, and the Charybdis of uncertainty ('an open 

mind is an empty mind')? 102 Scholars and practitioners have been very 

creative in proposing solutions, but, as Berts effectively demonstrates, the 

most prominent ones — assuming the worst, multiple advocacy, 

consolidation, Devil's advocacy (e.g., 'Team B' exercises), sanctions and 

incentives, and 'cognitive rehabilitation and methodological consciousness' 

- all face important obstacles or have serious defects. 103 Betts is certainly 

correct that intelligence failures are inevitable, and he rather pessimistically 

concludes that the inherent limits of intelligence assessment suggest that we 

are unlikely to realize more than marginal benefits from procedural 

innovation. 

Fatalism may be entirely appropriate - if we insist upon thinking of 

'better' intelligence as an improvement in some rate of success, or as a 

purely intellectual measure of the accuracy of an intelligence community's 

judgments of an adversary. But we may be able to improve the effectiveness 

of intelligence quite dramatically simply by keying on possible sources of 

intellectual error. If intelligence failures are inevitable, why not exploit that 

fact when attempting to serve policy? 

Because intelligence assessment is so heavily dependent upon 

assumptions and specific modes of inference, it stands to reason that if 

policy makers were more fully aware of them, they would be better able to 

read, understand, and make use of intelligence. National leaders are often 

sophisticated analysts in their own right, who operate on the basis of certain 

assumptions about the world, and who draw their own conclusions on the 

basis of their own beliefs. There is ample evidence of this on all sides in the 

Cuban missile crisis, certainly, although of the three leaders Khrushchev 

appears to have relied upon his own analysis most heavily, and Kennedy 

least. But all leaders are, to some extent, their own analysts, and are 

therefore confronted with the task of having to make sense of intelligence 

assessments which may or may not reinforce their own expectations. Some



political leaders - Castro, for example - have a fascination for intelligence, 

and like to wade into the assessment process, voraciously seeking details on 

the source and content of the information their intelligence services acquire. 

Most, perhaps, are too busy to do this even should they wish to do so, and 

must be content to receive the intelligence community's distilled wisdom. 

But relatively few, we suspect, think hard about their own assumptions, 

beliefs, and modes of inference, let alone those of their intelligence analysts, 

and accordingly most have but a dim appreciation of the underlying points 

of, and reasons for, similarity or disagreement. 

A political leader with strong beliefs and a high degree of selfconfidence 

may simply dismiss as flawed an intelligence report dissonant 

with his or her own judgments. Over time, given enough points of 

disagreement, this would quite naturally lead to an intelligence-policy rift. 

If intelligence analysts took pains, however, to flag the assumptions and 

inference patterns undergirding their estimates, this would not only tend to 

improve the rigor and consistency of estimates, but it would enable policy 

makers to see more easily why they agreed or disagreed. In some cases this 

might even lead to a productive dialogue between policy makers and 

analysts on crucial assumptions. This was precisely the rationale for an 

intriguing innovation instituted by the former Chairman of the National 

Intelligence Council, Joseph S. Nye Jr, who required intelligence estimates 

to identify and flag the crucial assumptions upon which they were based. 

The intelligence community initially resisted the requirement, but came to 

regard it as salutary for helping structure and clarify their own 

assessments. 104 For policy makers, it provided several points at which to 

engage an otherwise terse and remote summary document. 

As policy makers bear the ultimate burden of responsibility, anything 

that helps them to better understand both intelligence and the 

presuppositions of policy represents a potentially important contribution. 

We can imagine taking Nye's innovation two steps further: (1) offering 

incoming national leaders briefings on the nature and limits of intelligence 

early in their tenure; and (2) wherever possible appending to intelligence 

estimates analyses of how crucial judgments in it would change if certain 

key assumptions were modified or relaxed. 105 These practices would help 

prime policy makers to appreciate the performance limits of intelligence, 

and to integrate estimates more fully and more easily with their own 

judgments. At the same time, they would preserve intelligence estimates 

intact for those policy makers who are only interested in the distilled 

wisdom of the intelligence community and who have neither the time nor 

the inclination to peruse appendices. 

The main attractiveness of the practice lies in its potential for engaging 

decision makers in the intelligence-policy process. From a practical point of



view, it makes little sense to conclude, as did Dean Rusk, McGeorge Bundy, 

and John McCone, that US intelligence estimates 'took too little account of 

the possibility of a dangerous new departure in Soviet policy'. 106 What does 

it mean to 'take more account of a possibility'? Would this have materially 

affected President Kennedy's use of CIA estimates? We think not. However, 

even if the CIA had not changed its estimates, it might have been useful to 

try to make more explicit the assumptions on the basis of which they drew 

the conclusion that Khrushchev most likely would not deploy nuclear 

weapons to Cuba, and it might have been helpful to try to identify the 

conditions under which Khrushchev might attempt to do so (e.g., if he failed 

to understand that a US president would face enormous domestic and 

international pressure not to tolerate such a deployment; if he 

underestimated US intelligence capabilities; and if he were motivated by 

fear and insecurity, rather than by opportunity). At a minimum this would 

have given Kennedy and the ExComm a head start on trying to fathom 

Khrushchev's mistake after 15 October. 107 

But no adjustment to the presentation and content of estimates stands 

even a remote chance of improving intelligence performance when political 

leaders ignore, manipulate, lead, or censor their intelligence services. Both 

Amuchastegui and Fursenko and Naftali provide ample evidence of 

pathological intelligence-policy relationships caused by meddling or 

overconfident leaders. 108 This serves to underscore the importance of 

recalling that effective policy requires a certain disposition on the part of 

political leaders, as well as a certain competence on the part of professional 

intelligence analysts. The wide variation evident in the three intelligencepolicy 

relationships examined in this volume strikes us as utterly typical of 

intelligence-policy relationships in general. If so, this may be the only sense 

in which the Cuban missile crisis was run-of-the-mill. 

Certain events may reflect the significant dimensions of all your life, 

mirroring your entire history in a passing moment. Have you ever had 

an experience like that? Have you been caught by an event that 

suddenly pulled the curtains back? 

- Michael Murphy, Golf in the Kingdom



NOTES 

1. Throughout we will cite the essays in this volume by author, without title. 

2. Fursenko and Naftali, p.79; and Amuchastegui, p.89. 

3. Garthoff, p.45, citing U.S. Department of State, Foreign Relations of the United States, 

1961-1963, Vol. XI, Cuban Missile Crisis and Aftermath (Washington DC: GPO 1997) 

p.625 (hereafter cited as FRUS, 1961-63). 

4. Fursenko and Naftali, p.66. 

5. 'Before being sent out to the US to run the KGB's principal station there, Feklisov had 

overseen the intelligence assessment in 1960 that Eisenhower was unlikely to invade 

without a Cuban assault on Guantánamo Bay or the creation of a missile base on the island. 

Two years later, he still believed this, despite the change in administration and the 

experience of the Bay of Pigs. In the middle of March 1962, Feklisov reported that 

Kennedy was unlikely to approve an invasion because a military intervention would 

undermine the Alliance For Progress.' Fursenko and Naftali, pp.73-4. Surviving senior 

members of the Kennedy administration overwhelmingly concur that Kennedy would not 

have authorized an invasion of Cuba without some serious provocation, such as a 

deployment of Soviet nuclear missiles to Cuba James G. Blight and David A. Welch, On 

the Brink: Americans and Soviets Reexamine the Cuban Missile Crisis, 2nd ed. (NY: 

Noonday 1990); James G. Blight, Bruce J. Allyn, and David A. Welch, Cuba on the Brink: 

Castro, The Missile Crisis, and the Soviet Collapse (NY: Pantheon 1993) passim. 


7. Amuchastegui, pp.99-100. 

8. The weight of evidence and testimony available today would strongly suggest that 

Khrushchev did not seriously begin to consider a nuclear deployment to Cuba until April 

or May 1962, and that planning for the operation did not begin until shortly thereafter. See, 

e.g., Gen. Anatoli I. Gribkov and William Y. Smith, Operation Anadyr: U.S. and Soviet 

Generals Recount the Cuban Missile Crisis (Chicago: Edition Q 1994); Richard Ned 

Lebow and Janice Gross Stein, We All Lost the Cold War (Princeton UP 1994); Raymond 

L. Garthoff, Reflections on the Cuban Missile Crisis, rev. ed. (Washington DC: Brookings 

1989); Blight and Welch, On the Brink (note 5); Blight, Allyn, and Welch, Cuba on the 

Brink (note 5). The second reason which apparently led Cuban intelligence to conclude that 

Khrushchev must have conceived the deployment much earlier was the judgment that the 

Soviet Union - a notoriously sluggish and bureaucratic state - could not have planned and 

carried out such an ambitious redeployment in a mere 90 days. It is important to note, 

however, that Cuban intelligence apparently believed that the deployment was substantially 

complete by mid-September, and did not appreciate that the first Soviet MRBMs in Cuba 

only became operational in late October - four and a half months after planning could have 

begun in earnest following a late-May decision to proceed - and that the IRBMs never 

arrived. While this would still be a significant accomplishment for a notoriously sluggish 

and bureaucratic state, it could not be considered downright unbelievably rapid. 


10. Blight, Allyn, and Welch, Cuba on the Brink (note 5), pp.77-80. The second task of the 

delegation was to assess the prospect that strategic nuclear missiles could be deployed in 

Cuba without the United States discovering them. See p. 185, below. 

Fursenko and Naftali are correct to state that Alekseev 'was [not]consulted on the 

project' (p.75) if by this they mean that he was not brought in on the deliberations before 

Khrushchev had made up his mind to propose the deployment to Castro. However, as 

Alekseev recounts the tale, once Khrushchev had decided to make the offer, he did ask 

Alekseev how he thought Castro would react. Alekseev predicted incorrectly that Castro 

would not agree. Blight, Allyn, and Welch, Cuba on the Brink (note 5) pp.77-8. 

11. Timothy Naftali, personal communication. 

12. See, e.g., Graham T. Allison, Essence of Decision: Explaining the Cuban Missile Crisis 

(Boston: Little, Brown 1971) p. 110. 

13. Amuchastegui, p. 101. 

14. Since Cuban intelligence apparently discovered the true nature of the deployment in due



course, it is conceivable that they simply inferred that Soviet intelligence in Cuba must 

have been aware of it as well. This might have been an accurate judgment. But even if it 

were incorrect, it would be natural, because the Cubans - like the Americans - understood 

the Soviet Union as a coherent rational actor. See our discussion in note 96, below. 

15. Fursenko and Naftali, pp.80-82. 

16. Ibid. p.82. 

17. Fursenko and Naftali, p.70; Amuchastegui, pp.91-2. 

18. Nicola Miller, Soviet Relations with Latin America 1959-1987 (Cambridge UP 1989); 

Jorge I. Dominguez, To Make a World Safe for Revolution: Cuba's Foreign Policy 

(Cambridge, MA: Harvard UP 1989). 

19. Amuchastegui, p.99. 

20. Personal communications. Naftali adds in another personal communication: 'It seems odd 

that Alekseev would have been recalled [to Moscow] in the manner he was if Che was on 

his way to Cuba. [Also,] I have a clear recollection of the letters prepared by the Presidium 

for Castro in the first week of May, and not one refers to a fraternal visit by El Che.' 

21. Aleksandr Fursenko and Timothy Naftali, 'One Hell of a Gamble': Khrushchev, Castro and 

Kennedy, 1958-1964 (NY: Norton 1997) p.70. Amuchastegui finds it difficult to imagine 

that Che would have made such a probe at that time, and recalls that the chief purpose of 

Che's visit was to discuss economic cooperation (personal communication). 


23. Che's next documented voyage to the Soviet Union took place in late August 1962. The 

chief purpose of that trip was to propose making the deployment public. It seems unlikely 

that Amuchastegui is thinking of this particular trip, as increasing levels of military aid 

were not on the agenda. Blight and Welch, On the Brink (note 5) pp.333-4. 

24. Fursenko and Naftali, 'One Hell of a Gamble' (note 21) p.168. 

25. Ibid. p. 187. 

26. Garthoff, pp.28-9, 58, note 28. 

27. See ibid. p.28. 

28. Beth Fischer gives good reasons for preferring a cognitive to a motivational explanation in 

her essay [pp. 28-9]. 

29. Amuchastegui, pp.115-16, note 38. 

30. Garthoff, pp.22-3. 

31. See, e.g., the reports of 27 Sept. and 1 Oct., reprinted in Mary S. McAuliffe (ed.) CIA 

Documents on the Cuban Missile Crisis (Washington DC: CIA History Staff 1992) 

pp. 107-9 (hereafter CIA Documents). 

32. [Richard Lehman], Excerpt from Memo for Director of Central Intelligence, 'CIA 

Handling of the Soviet Buildup in Cuba', 14 Nov. 1962 [Excerpt], Top Secret, declassified; 

in CIA Documents (note 31) p.99. 

33. Certainly there is ample evidence that the KGB played at best a minor role throughout. It 

would appear from the somewhat thinner testimonial and documentary record on Soviet 

military intelligence that the GRU was marginalized as well. See generally Fursenko and 

Naftali, 'One Hell of a Gamble' (note 21) passim. 


35. In his annual message to Congress on 2 Dec. 1823, President James Monroe warned 

European states against any future attempts to colonize the Americas, declared that the 

United States would consider any attempt by the nations of Europe to extend their system 

into the hemisphere 'dangerous to our peace and safety', and disavowed American 

participation in European wars. See Ernest R. May, The Making of the Monroe Doctrine 

(Cambridg, MA: Belknap 1975). While Khrushchev may have believed that American 

participation in World Wars I and II represented a renunciation of the Monroe Doctrine, 

Troyanovsky would have known that no one in the United States so considered it. 

36. Blight, Allyn, and Welch, Cuba on the Brink (note 5) p.72. 

37. The KGB, despite a lack of highly-placed sources, seems to have had a fairly good 

understanding of President Kennedy, his attitude toward Cuba, and the constraints under 

which he operated domestically. Fursenko and Naftali, pp.68-9. 

The KGB's reading of the Bay of Pigs - that JFK had been pushed by the CIA and by



the hawks in his administration, and would, if left to his own devices, act less aggressively 

toward Castro (ibid, p.69) - is extremely interesting, and it would be useful to know exactly 

why the KGB drew this particular conclusion. Cf. James G. Blight and Peter Kornbluh 

(eds) Politics of Illusion: The Bay of Pigs Invasion Reexamined (Boulder, CO: Lynne 

Rienner 1997). 


39. Ibid., p.72. 

40. Ibid. 

41. It is conceivable that Khrushchev chose not to task the KGB with such an assessment in 

part because he feared it would compromise the secret, as Beth Fischer surmises (p. 165). 

There is no direct evidence to this effect, however. 


43. Blight and Welch, On the Brink (note 5) pp.333^4. It is interesting to note that senior 

members of the Kennedy administration unanimously agree that had Khrushchev 

attempted to deploy the missiles openly, the United States would have been in a much more 

difficult position. The secrecy and deception surrounding the deployment enabled the 

United States to characterize it as sinister and aggressive, and to deflect attention from the 

undeniable fact that what the Soviet Union and Cuba were doing was entirely legal. 

44. Ibid. p.239; Sergei Khrushchev, personal communication. 

45. Graham Allison has suggested that organizational routines may explain the failure of the 

Soviet military to camouflage the missile sites effectively; Essence of Decision, (note 12), 

p.111. An explanation more consistent with the available evidence is incompetence. See 

David A. Welch, 'The Organizational Process and Bureaucratic Politics Paradigms: 

Retrospect and Prospect', International Security Mil (Fall 1992) pp.112-46. 

46. 'Sharaf Rashidov had reported to the Defense Council that Cuba's forests would provide 

just the needed cover for our missiles. Only someone with absolutely no competence in 

such technical matters could have reached such a conclusion.' Gribkov and Smith, 

Operation Anadyr (note 8), p.40. 



49. Ibid. p.102. 

50. Ibid. pp.104-5. 

51. Fursenko and Naftali, 'One Hell of a Gamble' (note 21) pp.268, 272. 

52. Castro has spoken in general terms about his conviction that the United States, once ready, 

would pounce, and would seek to do so without warning. Evidently he was determined not 

to make the mistake Stalin had made in 1941 of failing to prepare to meet an assault that 

could be launched at any time. See Blight, Allyn and Welch, Cuba on the Brink (note 5) 

pp.109-10. 

53. US Senate, Committee on Armed Services, Investigation of the Preparedness Program, 

Interim Report on the Cuban Military Buildup by the Preparedness Investigating 

Subcommittee (Washington DC: GPO 1963) (hereafter Stennis Report); President's 

Foreign Intelligence Advisory Board, James R. Killian Jr, Chairman, Memo for the 

President and Report, 4 Feb. 1963, Top Secret, 10 pp.; reprinted in CIA Documents (note 

31) pp.361-71. 

54. McCone, Memo for the President, 28 Feb. 1963, and 'Conclusions', in ibid, pp.373-6. 

55. Richard K. Berts, 'Analysis, War, and Decision: Why Intelligence Failures are Inevitable', 

World Politics 31/1 (Oct. 1978) pp.61-89. Many of the constraints on intelligence 

assessment that Berts identifies may be understood in terms of the psychological dynamics 

. Fischer discusses in her essay, although his study predates much of the research upon 

which Fischer draws. Betts concludes that '[intelligence failure is political and 

psychological more often than organizational'. Ibid. p.61. 

Betts also draws our attention to the fact that intelligence communities are often engaged 

in a zero-sum strategic interaction, whereby success for one logically implies failure for the 

other. The fact that even in the most celebrated cases surprise is never truly absolute further 

underscores the sense in which 'the best possible' performance is always something less 

than 'perfect performance'. Richard K. Betts, 'Surprise Despite Warning: Why Sudden



Attacks Succeed', Political Science Quarterly 95/4 (Winter 1980-81) pp.551-72. See also 

Michael Handel, 'Intelligence and the Problem of Strategic Surprise', Journal of Strategic 

Studies 7/3 (Sept. 1984) p.230. 

56. See, e.g., Garthoff, p.23: 'Some ambiguous or equivocal intelligence indications in 

September-October were noted but did not command as much attention as, in hindsight, 

they should have. In particular, it was observed that several Soviet merchant ships with 

large hatches en route to Cuba were riding high in the water, indicating bulky but relatively 

light cargo, such as missiles. Some of these ships were seen unloading normal cargoes, but 

others were reported to be unloaded at night, again a possible indicator of a cargo of special 

sensitivity. After the missiles were discovered, in retrospect it was recognized that these 

particular ships had undoubtedly brought in the missiles in early to mid-September.' US 

intelligence did notice all of this at the time, but it may be doubted whether the CIA had 

compelling reasons in 1962 to regard it all as more significant than they did. 

57. The classic study is Roberta Wohlstetter, Pearl Harbor: Warning and Decision (Stanford 

UP 1962). For an early yet insightful comparison of the Cuban and Pearl Harbor cases, see 

idem, 'Cuba and Pearl Harbor: Hindsight and Foresight', Foreign Affairs 43/4 (July 1965) 

pp.691-707. 

58. Monday-morning quarterbacking is much easier for football fans than for critics of 

intelligence, because in football it is possible to play the percentages. We can justifiably 

criticize a coach for calling a long pass on third-and-inches, because a running play would 

have a much better chance of success (and a much lower chance of resulting in a turnover). 

Critics of intelligence do not ordinarily have good-quality base rate information with which 

to second-guess intelligence professionals. 

59. Klaus Knorr, 'Failures in National Intelligence Estimates: The Case of the Cuban Missiles', 

World Politics 16/3 (April 1964) pp.455-67, 460. 

60. Special National Intelligence Estimate (SNIE) 85-3-62, The Military Buildup in Cuba, 19 

Sept. 1962, Secret; in CIA Documents (note 31), p. 93; also reprinted FRUS. 1961-1963 

(note 3), Vol.X, Cuba, Jan.l961-Sept. 1962, pp.1070-80. 

61. Garthoff notes that some of the authors of the September SNIE subsequently argued that 

the estimate had been correct and that Khrushchev had been mistaken (p.21). There is a 

sense in which this is undoubtedly true. One possible rebuttal is that the CIA should have 

known that Khrushchev would not consult his American specialists, and that they should 

have factored this into their estimate. This reply, however, begs all of the difficult questions 

to which we allude under the previous head. 

62. Echoing the point of the previous paragraph, it is worth noting that had President Kennedy 

not chosen to deliberate for a week before choosing his response to the Soviet deployment, 

he might very well have done as Cuban intelligence anticipated. On 16 Oct., the mood in 

the White House cabinet room was decidedly belligerent. The delay was crucial for 

permitting tempers to cool. See generally James G. Blight, The Shattered Crystal Ball: 

Fear and Learning in the Cuban Missile Crisis (Savage, MD:Rowman & Littlefield 1990). 

Although the timing of the American response was not something Cuban intelligence 

controlled, it might well have had a significant effect on our retrospective assessment of 

their performance. 

63. Amuchastegui, p.114, note 26. 

64. According to Amuchastegui, DGI also considered an American attack inevitable (ibid.). 

Since this was an open-ended prediction, we are technically still not in a position to 

determine whether that estimate was correct. 

65. Stennis Report (note 53), p.2. 

66. Ibid. p. 11. 

67. Garthoff, p.21. Note that Garthoff does not make this statement specifically in reply to the 

Stennis Report. 

68. Ibid., p.23. 

69. Stennis Report (note 53), p.5. Cf. Amuchastegui's claim that '[n]othing [Cuban sources] 

said would be credited unless they showed up in Miami with a Soviet missile' (p.103). 

70. See pp.181-2 above. 

71. Fischer's essay extensively documents the propensity to evaluate information in the light



of pre-existing beliefs. Clearly the large number of false reports of missile sightings by 

Cuban agents and refugees had led CIA analysts to believe that Cuban sources were 

unreliable, and that their reports had ulterior purposes. 

72. It might be replied here that certain kinds of behavior unambiguously telegraph intentions. 

For example, we can infer reliably that a leader who masses troops on the border in a full 

battle-ready posture intends to attack. This behavior, however, is fully consistent with other 

intentions: e.g., to provoke an attack, to intimidate, to signal acute concern, and so forth. 

See Michael Handel, Perception, Deception and Surprise: The Case of the Yom Kippur War 

(Jerusalem: Leonard Davis Institute 1976); and Janice Gross Stein, 'Calculation, 

Miscalculation, and Conventional Deterrence II: The View from Jerusalem', in Robert 

Jervis, Richard Ned Lebow and Janice Gross Stein (eds) Psychology & Deterrence 

(Baltimore, MD: Johns Hopkins UP 1985) pp.60-88. 

73. Fischer, p. 159. 

74. 'Failures in National Intelligence Estimates' (note 59), p.462, 

75. Ibid. pp.463, 465. 

76. The best example is Arnold L. Horelick, 'The Cuban Missile Crisis: An Analysis of Soviet 

Calculations and Behavior', World Politics 16/3 (April 1964) pp.363-89. Cf. also Arnold 

Horelick and Myron Rush, Strategic Power and Soviet Foreign Policy (U. of Chicago Press 

1966) p.141: 'Soviet calculations throughout the Cuban missile crisis can be reconstructed 

with some degree of confidence, because the confrontation with the United States was so 

open'. 

77. Garthoff, pp.24-6, 36, 46-7. 

78. Similarly, US analysts and policy makers had insufficient evidence on the basis of which 

to draw reliable conclusions about what Soviet leaders might be willing to consider by way 

of retaliation for American military action against Cuba. Garthoff writes: 'In response to 

either an air strike or an invasion, the intelligence estimators concluded that because the 

Soviet leaders "would almost certainly not resort to general war and could not hope to 

prevail locally, we believe that the Soviets would consider retaliatory actions outside 

Cuba", and the most likely place would be Berlin - but "major harassments", not a military 

move on West Berlin. As for a countering strike against some US base around the world 

(such as the missiles in Turkey, although not specified), a Soviet retaliatory move of this 

kind was considered "possible but in our view unlikely". Again, this judgment was never 

tested, but although the estimated Soviet response (to "consider" retaliation by a noncombat 

harassing action) was much more mild than many (including Secretary McNamara 

and some other members of the ExComm) believed likely, and was certainly a prudent 

caution, it appears from all we now know to have been the most the Soviet leaders would 

have done, if that.' Ibid. pp.30-1. 

79. Ibid. p.21. N.B. that Garthoff argues, on the basis of present knowledge, that this is what 

the estimate should have concluded, 

80. Klaus Knorr therefore seems to us to be splitting hairs unduly finely when he states, 'The 

question is: why did the intelligence community fail to warn the government earlier that 

such a Soviet move was distinctly possible, if not probable, instead of estimating that it was 

improbable, though not impossible?' 'Failures in National Intelligence Estimates' (note 

59), p.456. 

81. Garthoff, p.18. We would choose some word other than 'compel', in view of the fact that 

the United States agreed secretly to withdraw its Jupiter missiles from Turkey as a quid pro 

quo. 

82. Ibid. p.24. 

83. See, e.g., Mark Kramer, 'The Cuban Missile Crisis and Nuclear Proliferation', Security 

Studies 5/1 (Autumn 1995) pp.171-9 ; and James G. Blight and David A. Welch, 'On 

Historical Judgment and Inference: A Reply to Mark Kramer', ibid. 5/4 (Summer 1996) 

pp.172-82. We should qualify this last statement by noting that we do not know whether 

US signals intelligence had collected anything that would have clarified any of these 

matters. 

84. Garthoff, pp.27-9. 

85. We appreciate that how one bounds the relevant period may be a choice of some



consequence. Defining 'an event' spatio-temporally is, to some extent, arbitrary, and is 

perhaps usefully understood as a function of socio-cultural propensities. We have remarked 

before upon the different understandings of the crisis in the United States, the Soviet Union 

(now Russia), and Cuba, and the significance of the different names by which it is known 

in each country. We have also argued that culturally-rooted differences in problemrepresentation 

help account for its genesis and greatly complicated its resolution. See 

Blight, Allyn, and Welch, Cuba on the Brink (note 5) pp.318-72. We do not believe that 

this represents a serious obstacle for the present point, however: as long as it is possible to 

conceptualize an intelligence encounter as having taken place over some period of time, 

and as consisting of some number of discrete phases or episodes, it should be possible — 

within limits - to arrive at some evaluation of performance, even if those 

conceptualizations are arbitrary, and the episodes in fact linked. Thus by contrasting US 

intelligence performance with respect to Cuba over the period 1958-61 and 1961-62 — 

with the Bay of Pigs and the Cuban missile crisis as foci - we can meaningfully conclude 

that the CIA performed poorly in one period and well in another, and we can learn 

something of practical consequence from the contrast. 

86. Plato, Meno, 80d; in Five Dialogues, trans. G.M.A. Grube (Indianapolis: Hackett 1981) p. 

69. 

87. Berts, 'Analysis, War, and Decision' (note 55), p.61. 

88. Consider, for example, the following: 'Castro in March had moved to purge key old 

Moscow-line communist leaders from his regime, and some in the US intelligence 

community saw this as causing a significant rift in Cuban-Soviet relations. Military 

shipments from the USSR, after a buildup in 1961, were lower in the first half of 1962. The 

evidence before July 1962 thus tended to support an estimate that a major Soviet military 

buildup in Cuba was not likely.' Garthoff, p.22. US intelligence apparently inferred from 

the Escalante affair that Soviet-Cuban relations would deteriorate; yet, as Amuchastegui 

explains, the purpose of Castro's purge was precisely the reverse: to prepare the ground for 

a closer security relationship (Amuchastegui, p.92). The purge itself was consistent with 

diametrically opposite motivations, and US intelligence simply leapt to the conclusion that 

it implied one in particular. Garthoff considers the US judgment 'intuitive', but this, of 

course, is merely another way of saying that it was more consistent with suppressed 

assumptions. 

89. This is a complicated claim whose full exegesis would take us very far afield. The gist, 

however, is that leaders of states make decisions on the basis of idiosyncratic beliefs and 

judgments about the lessons of history and how the world works, and do so consciously in 

reaction to the historical record, often precisely so as to surprise. Important choices are 

often unprecedented, or have important unique features, such that the historical record 

provides no basis for estimating a probability distribution. Among the important works that 

touch various aspects of this issue are Robert Jervis, 'The Future of World Politics: Will it 

Resemble the Past?' International Security 16/3 (Winter 1991/92) pp.39-73; Yuen Foong 

Khong, Analogies at War: Korea, Munich, Dien Bien Phu, and the Vietnam Decisions of 

1965 (Princeton UP 1992); and Richard E. Neustadt and Ernest R. May, Thinking in Time: 

The Uses of History for Decision-Makers (NY: Free Press 1986). See also Masato Kimura 

and David A. Welch, 'Specifying "Interests": Japan's Claim to the Northern Territories and 

Its Implications for International Relations Theory', International Studies Quarterly 42/2 

(June 1998) pp.213-44. 

90. Among political scientists there is a widespread impression that 'expected utility theory' 

can serve as the basis for intelligence assessment. Bruce Bueno de Mesquita claims a 

successful point-prediction rate of over 90 per cent across a variety of political decisionmaking 

domains for a particular expected-utility model based upon Black's median voter 

theorem. Bruce Bueno de Mesquita, David Newman and Alvin Rabushka, Forecasting 

Political Events: The Future of Hong Kong (New Haven, CT: Yale UP 1985); Duncan 

Black, Voting in Committees and Elections (Cambridge UP 1958). Bueno de Mesquita's 

model is not, however, an 'expected utility' theory at all, merely a computational device for 

aggregating subjective judgments of actors' weighted preferences on a given issue. Its 

predictive power - whatever that may be (many of the details are classified) - depends



essentially upon the quality of the inputs, which are themselves generated by area 

specialists in a non-scientific way. For further discussion, see Janice Gross Stein and David 

A. Welch, 'Rational and Psychological Approaches to the Study of International Conflict: 

Comparative Strengths and Weaknesses', in Alex Mintz and Nehemia Geva (eds) 

Decision-Making on War and Peace: The Cognitive-Rational Debate (Boulder, CO: Lynne 

Rienner 1997) pp.51-77. 

91. Imre Lakatos, 'Falsification and the Methodology of Scientific Research Programmes', in 

idem and Alan Musgrave (eds) Criticism and the Growth of Knowledge (Cambridge UP 

1970) pp.91-196; idem The Methodology of Scientific Research Programmes, Vol.1 (ibid. 

1978). To simplify a bit, Lakatos argued that we should prefer theory x to theory y when x 

explains everything y explains, plus additional facts. 

92. Betts, 'Analysis, War, and Decision' (note 55), p.63. 

93. 'I know what you want to say, Meno. Do you realize what a debater's argument you are 

bringing up, that a man cannot search either for what he knows or for what he does not 

know? He cannot search for what he knows - since he knows it, there is no need to search 

- nor for what he does know, for he does not know what to look for.' Meno, 80e, p.69. 

94. Ibid. 81b-d, pp.6-70. 

95. Hans-Georg Gadamer, Truth and Method, trans. Garrett Barden and John Cumming 

(London: Sheed & Ward 1975); Willard Van Orman Quine, Word and Object (Cambridge, 

MA: MIT Press 1964); Noam Chomsky, Syntactic Structures (The Hague: Mouton 1965). 

See also Graeme Nicholson, Seeing and Reading (London: Macmillan 1984). 

96. For applications of a competing perspective (which the author calls 'critical 

constructivism') on the phenomenon as it bears on problem-representation in the Cuban 

missile crisis, see Jutta Weldes, Constructing National Interests: The U.S. and the Cuban 

Missile Crisis (U. of Minnesota Press forthcoming 1998). For a critical cultural perspective 

as it bears on intelligence assessment, see Robert B. Bathurst, Intelligence and the Mirror: 

On Creating an Enemy (Oslo: PRIO 1993). 

It would certainly appear that what might be thought of as 'intellectual culture' has a 

profound effect on political judgments of all kinds, including intelligence assessments. In 

the Western world, at least five stock assumptions about how international politics works, 

and about how leaders make decisions — roughly congruent with the 'Realist' 

understanding of international politics - tend to dominate. Among the most important of 

these assumptions are (1) that all states always seek 'power'; (2) that leaders are 'rational'; 

(3) that all leaders understand, and respond to, threats and incentives; (4) that the 

adversary's behavior is always intentional; and (5) that the adversary's actions are always 

coordinated and meaningful. Typically, when leaders and analysts attempt to make sense of 

an adversary's behavior, they infer evil intent from actions they dislike, and dismiss as 

disingenuous and instrumental arguments adversaries make that appeal to notions such as 

justice or fairness. Often these are important mistakes. These assumptions exert a powerful 

spin on ambiguous data, and can lead to tragic misjudgments. For extensive discussion, see 

David A. Welch, Justice and the Genesis of War, Cambridge Studies in International 

Relations 29 (Cambridge UP 1993); and idem, 'Remember the Falklands? Missed Lessons 

of a Misunderstood War', InternationalJournal 52/3 (Summer 1997) pp.483-507. 

97. Garthoff, p.46. 

98. As we suggest in our introductory essay, these may not have been effective in the absence 

of credible offers to alleviate Khrushchev's anxieties, and it is difficult to imagine what 

those offers might have been, given the political realities of the day. Moreover, this is not 

an intelligence failure: Kennedy did not need the impetus of an intelligence assessment to 

communicate more clearly with Khrushchev. Yet, conceivably, had the CIA raised the 

alarm earlier, it might have been able to serve policy even better than it did. 

99. There were glimmers of understanding: 'In seeking to understand why Khrushchev and the 

other Soviet leaders believed they could succeed', Garthoff writes, 'the ORR study 

concluded (in my view correctly) that Khrushchev, "having lived restively under the 

shadow of U.S. strategic bases for more than a decade", and now having his first 

opportunity to do likewise, "may have felt confident that the U.S. would understand the 

rules as he did - that military bases on the opponent's periphery are facts of great power



life which fall far short of a provocation to war".' Ibid.,p.39. But as Garthoff also notes 

extensively, even after the shock of the missile crisis, retrospective evaluations consistently 

underestimated Khrushchev's deterrent and defensive motivations. 

100. It seems to us undeniable that the self-awareness of the US government - perhaps of the 

entire country - was so deficient, and so discrepant with Cuban perceptions of the United 

States, as to be laughable were not the stakes too high for laughter. In all the posturing, no 

one seems to have asked two questions: (1) Are our actions scaring the Cubans as well as 

deterring the Soviets? (2) If so, might Cuba's sense of insecurity drive the Soviets to 

undertake desperate measures on its behalf? From the American perspective, Cuba was 

simply absent from the Cuban missile crisis. 

While it is possible to imagine that some American official or commentator might have 

posed these questions and answered them affirmatively in 1962, it is difficult to imagine 

this having had much effect on America's Cuba policy, given the grip that the rational 

deterrence/Cold War perspective had on the American imagination at the time. 

101. See, e.g., Janice Gross Stein, 'Political Learning by Doing: Gorbachev as Uncommitted 

Thinker and Motivated Learner', International Organization 48/2 (Spring 1994) 

pp. 155-83. Notwithstanding their judgment that the United States would eventually try to 

destroy the Cuban Revolution, Cuban intelligence concluded that an American attack was 

not imminent after the Bay of Pigs because (a) the pattern of activity conducted under the 

rubric of Operation 'Mongoose' seemed ill-designed to prepare the ground for an 

American invasion; and (b) Cuban intelligence believed that the United States would not 

invade Cuba without adequate diplomatic 'cover' and the sanction of other Latin American 

states. Amuchastegui, pp.96-7. Clearly DGI drew heavily upon assumptions about 

American behavior when attempting to predict American behavior. We disagree with 

Fischer that '[d]ata-driven thinking could have led [Khrushchev] to the (correct) 

conclusion that the Kennedy administration was not planning an attack in 1962.' Fischer, 

p.156. Since American behavior was consistent with different interpretations, data-driven 

thinking, even if it were possible, would have led to no particular conclusion at all. Instead, 

we would suggest that Khrushchev would have reached the same conclusion as Cuban 

intelligence if he had relied upon a different set of assumptions with which to interpret 

ambiguous American behavior. 

Fischer also writes: 'It is also interesting to note that while Cuban and American 

intelligence reached the same (inaccurate) conclusion about Soviet motives, they appear to 

have done so via different methods of information processing. American analysis appears 

to have been heavily theory driven. Beginning with the assumption that the Soviet Union 

was a cautious, opportunistic aggressor unlikely to risk a nuclear deployment to Cuba, 

American analysts concluded that the purpose of the deployment was to give the Soviets a 

strategic advantage over the United States once the deployment became evident. Cuban 

intelligence, on the other hand, inferred Soviet motives from the available data — 

specifically, the discrepancy between the dire Soviet assessments of the American threat 

and Cuban assessments that such a threat was receding. This finding suggests that there is 

no necessary relationship between one method of information processing and the accuracy 

of inferences. Data-driven and theory-driven information processing can both lead to 

inaccurate assessments.' Ibid. pp.163-4. We understand Cuban judgments here as no less 

theory-driven than American judgments. 

According to Amuchastegui's account, Cubans had been somewhat suspicious of Soviet 

intentions all along, and therefore, among the available set of working assumptions about 

Soviet behavior, was the hypothesis that Soviet policy with respect to Cuba was driven by 

narrow self-interest rather than altruism. Cuban intelligence was never wedded to the 

altruism assumption and had no difficulty assimilating the available evidence to an 

alternative assumption. Note also that Soviet dissimulation did not logically imply that the 

Soviet motive in deploying nuclear missiles to Cuba was to gain a strategic advantage over 

the United States. Many possible purposes could be served thereby, as Americans 

recognized — including the defense of Cuba (Khrushchev may simply have believed that the 

Cubans would never take what was necessary for their defense unless he exaggerated the 

American threat). See., e.g, Allison, Essence of Decision (note 12) pp.43-56.



102. Betts, 'Analysis, War, and Decision' (note 55) p.84. 

103. Ibid. pp.73-84. 

104. We are not clear on the chief reason for the resistance, though it is possible to imagine that 

estimators feared it would be onerous, distracting, or superfluous. 

105. One of the striking features both of US intelligence assessment and White House 

decisionmaking during the Cuban missile crisis was the partial success achieved in relaxing 

certain common assumptions and articulating the implications - specifically, the 

assumptions that the adversary is a 'unitary' and 'rational' actor. Garthoff writes that 'Two 

special national intelligence estimates issued on 19 October (SNIE 11-18-62) and 20 

October (SNIE-11-19-62) ... [recognized] "the possibility that the Soviets, under pressure 

to respond, would again miscalculate and respond in a way which, through a series of 

actions and reactions, could escalate into general war",' and that 'in any use of military 

force by the United States, the Soviet leaders would be "alarmed and agitated" (SNIE 11- 

18-62) or at least "surprised and probably alarmed" (SNIE-11-19-62).' Garthoff, p.30. US 

policy makers, similarly, 'prudently allowed for the possibility that even if Moscow 

instructions prohibited initiative of Soviet forces in Cuba to fire the nuclear missiles there, 

in case of an American air strike or invasion any surviving Soviet missiles might be fired 

either with or without authorization from Moscow. Both the general assumption of tight 

Moscow control, and the recognition that it nonetheless might fail, were properly taken into 

account.' Ibid. p.54. In our view, US policy makers were encouraged to relax these 

assumptions about Soviet behavior in part because of their own experience during the crisis 

- and particularly at its climax - of American command-and-control problems. See 

generally Blight and Welch, On the Brink (note 5), passim. Contrast Argentine leaders' 

rigid (and illogical) ascriptions of unity and rationality to British actions prior to and during 

the Falklands/Malvinas war of 1982 (Welch, 'Remember the Falklands?' note 96); Alberto 

A. De Vita, Malvinas 82: Como y Por Que (Buenos Aires: Institute de Publicaciones 

Navales 1994). 

106. McCone meeting with Kennedy, 5 Jan. 1963, in FRUS, 196-1963, Vol.XI, pp.702-4; cited 

by Garthoff, p.46. 

107. Note also that it would have helped minimize the dangers of post-hoc rationalization 

characteristic of several prominent retrospective studies that attempted to make sense of 

Khrushchev's calculations. See the discussion on pp.196-7, above. 

108. Amuchastegui, for example, writes: 'Not only did Cuban leaders privilege Soviet over 

Cuban assessments, but Cuban intelligence was instructed to treat the inevitable 

confrontation with the United States as its guiding strategic assumption shaping every 

aspect of data collection and interpretation' (p.98), and that 'in June 1962, Piñeiro had 

given strict instructions not to reflect upon the topic of Soviet forces and missiles in Cuba 

in any Cuban assessments' (p.100). 

109. Michael Murphy, Golf in the Kingdom (NY: Arkana 1972) p.127.

The Cuban missile crisis and intelligence performance

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?